Fix CI: resolve clang-tidy findings in the stream position tests

#5344 added two lines to unit-deserialization.cpp that clang-tidy
reports: modernize-return-braced-init-list for the remaining() helper
and readability-isolate-declaration for "json j1, j2, j3;".

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

.github/CONTRIBUTING.md

@@ -158,6 +158,15 @@ make amalgamate
 Running `make amalgamate` will also apply automatic formatting to the source files using
 [`Artistic Style`](https://astyle.sourceforge.net/). This formatting may modify your source files in-place. Be certain to review and commit any changes to avoid unintended formatting diffs in commits.
 
+If you add, rename, or remove a header in `include/nlohmann`, also regenerate the header list in
+[`BUILD.bazel`](https://github.com/nlohmann/json/blob/develop/BUILD.bazel) (requires CMake) by executing:
+
+```shell
+make BUILD.bazel
+```
+
+The amalgamation check in CI fails if any of these generated files is out of date.
+
 ## Recommended documentation
 
 - The library’s [README file](https://github.com/nlohmann/json/blob/master/README.md) is an excellent starting point to

.github/PULL_REQUEST_TEMPLATE.md

@@ -2,6 +2,7 @@
 
 - [ ] The changes are described in detail, both the what and why.
 - [ ] If applicable, an [existing issue](https://github.com/nlohmann/json/issues) is referenced.
+- [ ] If applicable, a fixed [OSS-Fuzz](https://issues.oss-fuzz.com) issue is referenced as `OSS-Fuzz: <id>` (see [fuzz testing](https://github.com/nlohmann/json/blob/develop/tests/fuzzing.md#handling-oss-fuzz-reports)).
 - [ ] The [Code coverage](https://coveralls.io/github/nlohmann/json) remained at 100%. A test case for every new line of code.
 - [ ] If applicable, the [documentation](https://json.nlohmann.me) is updated.
 - [ ] The source code is amalgamated by running `make amalgamate`.

.github/labeler.yml

@@ -29,6 +29,27 @@ labels:
   files:
     - ".github/external_ci/.*"
 
+- label: "CI"
+  files:
+    - ".github/(dependabot|labeler)\\.yml"
+
+- label: "aspect: binary formats"
+  files:
+    - "include/nlohmann/detail/input/binary_reader\\.hpp"
+    - "include/nlohmann/detail/output/binary_writer\\.hpp"
+    - "tests/src/unit-(bson|cbor|msgpack|ubjson|bjdata|binary_formats)"
+    - "tests/src/fuzzer-parse_(bson|cbor|msgpack|ubjson|bjdata)"
+    - "docs/mkdocs/docs/features/binary_formats/"
+    - "docs/mkdocs/docs/(api/basic_json|examples)/(to|from)_(bson|cbor|msgpack|ubjson|bjdata)"
+
+- label: "aspect: binary formats"
+  title: "(?i)(bson|cbor|msgpack|messagepack|ubjson|bjdata|binary format)"
+
+- label: "python"
+  files:
+    - "\\.py$"
+    - "requirements[^/]*\\.txt$"
+
 - label: "S"
   size-below: 10
 - label: "M"

.github/workflows/check_amalgamation.yml

@@ -57,13 +57,16 @@ jobs:
           python3 -mvenv venv
           venv/bin/pip3 install -r $MAIN_DIR/tools/astyle/requirements.txt
 
-      - name: Regenerate amalgamation and formatting
+      - name: Regenerate amalgamation, formatting, and BUILD.bazel
         run: |
           cd $MAIN_DIR
 
           python3 $TOOL_DIR/amalgamate.py -c $TOOL_DIR/config_json.json -s .
           python3 $TOOL_DIR/amalgamate.py -c $TOOL_DIR/config_json_fwd.json -s .
 
+          # the header list of the Bazel "json" target must match the files in include/
+          cmake -P cmake/scripts/gen_bazel_build_file.cmake
+
           ${{ github.workspace }}/venv/bin/astyle --project=tools/astyle/.astylerc --suffix=none --quiet \
             $INCLUDE_DIR/json.hpp $INCLUDE_DIR/json_fwd.hpp
 
@@ -87,7 +90,7 @@ jobs:
           mkdir -p ${{ github.workspace }}/patch
           git diff --patch --no-color > ${{ github.workspace }}/patch/amalgamation.patch
           if [ -s ${{ github.workspace }}/patch/amalgamation.patch ]; then
-            echo "The source code has not been amalgamated/formatted correctly. Diff:"
+            echo "The source code has not been amalgamated/formatted correctly or BUILD.bazel is out of date. Diff:"
             cat ${{ github.workspace }}/patch/amalgamation.patch
             echo "has_diff=true" >> "$GITHUB_OUTPUT"
           else

.github/workflows/comment_check_amalgamation.yml

@@ -95,13 +95,13 @@ jobs:
               issue_number: issue_number,
               owner: context.repo.owner,
               repo: context.repo.repo,
-              body: '## 🔴 Amalgamation check failed! 🔴\nThe source code has not been amalgamated and/or formatted correctly.'
+              body: '## 🔴 Amalgamation check failed! 🔴\nThe source code has not been amalgamated and/or formatted correctly, or `BUILD.bazel` is out of date.'
                     + (hasPatch ? '\n\n📎 A ready-to-apply patch is attached to the [failed workflow run](' + runUrl + ') as the `amalgamation-patch` artifact.'
                                   + ' Download it, then apply it locally from the repository root with:'
                                   + '\n\n```shell\ngit apply amalgamation.patch\n```\n\n'
                                   + 'This does not require installing astyle yourself.'
                                 : '')
                     + (first ? '\n\n@' + author + ' Please read and follow the [Contribution Guidelines]'
-                               + '(https://github.com/nlohmann/json/blob/develop/.github/CONTRIBUTING.md#files-to-change).'
+                               + '(https://github.com/nlohmann/json/blob/develop/.github/CONTRIBUTING.md#amalgamate-the-source-code).'
                              : '')
               })

.github/workflows/ubuntu.yml

@@ -100,7 +100,7 @@ jobs:
     container: ubuntu:focal
     strategy:
       matrix:
-        target: [ci_cmake_flags, ci_test_diagnostics, ci_test_diagnostic_positions, ci_test_noexceptions, ci_test_noimplicitconversions, ci_test_legacycomparison, ci_test_noglobaludls, ci_test_disableenumserialization, ci_test_skiplibraryversioncheck, ci_test_simdutf, ci_test_strict_nul_handling]
+        target: [ci_cmake_flags, ci_test_diagnostics, ci_test_diagnostic_positions, ci_test_noexceptions, ci_test_noimplicitconversions, ci_test_legacycomparison, ci_test_noglobaludls, ci_test_disableenumserialization, ci_test_skiplibraryversioncheck, ci_test_simdutf, ci_test_strict_nul_handling, ci_test_no_thread_local]
     steps:
       - name: Install build-essential
         run: apt-get update ; apt-get install -y build-essential unzip wget git libssl-dev

.github/workflows/windows.yml

@@ -124,11 +124,11 @@ jobs:
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
       - name: Run CMake (Release)
-        run: cmake -S . -B build -G "Visual Studio 17 2022" -A ARM64 -DJSON_BuildTests=On -DCMAKE_CXX_FLAGS="/W4 /WX"
+        run: cmake -S . -B build -G "Visual Studio 18 2026" -A ARM64 -DJSON_BuildTests=On -DCMAKE_CXX_FLAGS="/W4 /WX"
         if: matrix.build_type == 'Release'
         shell: pwsh
       - name: Run CMake (Debug)
-        run: cmake -S . -B build -G "Visual Studio 17 2022" -A ARM64 -DJSON_BuildTests=On -DJSON_FastTests=ON -DCMAKE_CXX_FLAGS="/W4 /WX"
+        run: cmake -S . -B build -G "Visual Studio 18 2026" -A ARM64 -DJSON_BuildTests=On -DJSON_FastTests=ON -DCMAKE_CXX_FLAGS="/W4 /WX"
         if: matrix.build_type == 'Debug'
         shell: pwsh
       - name: Build
@@ -158,6 +158,10 @@ jobs:
       # to fit: IMAGE_REL_AMD64_SECREL against `.debug_line'" because the
       # MinGW linker cannot relocate the debug sections this test produces.
       # The tests are only built and run here, so the debug info is not used.
+      # Do not add -O1 here to shrink the objects further: it does make them
+      # link, but the binaries clang 11.0.1 and clang 18.1.8 then produce crash
+      # before doctest prints its first line - 39 of 102 tests on clang 18.
+      # Keep the objects small by splitting the test files instead.
       - name: Run CMake
         run: cmake -S . -B build ^
              -DCMAKE_CXX_COMPILER="C:/Program Files/LLVM/bin/clang++.exe" ^

BUILD.bazel

@@ -30,8 +30,10 @@ cc_library(
         "include/nlohmann/detail/input/input_adapters.hpp",
         "include/nlohmann/detail/input/json_sax.hpp",
         "include/nlohmann/detail/input/lexer.hpp",
+        "include/nlohmann/detail/input/number_parse.hpp",
         "include/nlohmann/detail/input/parser.hpp",
         "include/nlohmann/detail/input/position_t.hpp",
+        "include/nlohmann/detail/input/string_scan.hpp",
         "include/nlohmann/detail/iterators/internal_iterator.hpp",
         "include/nlohmann/detail/iterators/iter_impl.hpp",
         "include/nlohmann/detail/iterators/iteration_proxy.hpp",
@@ -49,12 +51,14 @@ cc_library(
         "include/nlohmann/detail/meta/detected.hpp",
         "include/nlohmann/detail/meta/identity_tag.hpp",
         "include/nlohmann/detail/meta/is_sax.hpp",
+        "include/nlohmann/detail/meta/logic.hpp",
         "include/nlohmann/detail/meta/std_fs.hpp",
         "include/nlohmann/detail/meta/type_traits.hpp",
         "include/nlohmann/detail/meta/void_t.hpp",
         "include/nlohmann/detail/output/binary_writer.hpp",
         "include/nlohmann/detail/output/output_adapters.hpp",
         "include/nlohmann/detail/output/serializer.hpp",
+        "include/nlohmann/detail/recursion_depth_limit.hpp",
         "include/nlohmann/detail/string_concat.hpp",
         "include/nlohmann/detail/string_escape.hpp",
         "include/nlohmann/detail/string_utils.hpp",
@@ -67,7 +71,6 @@ cc_library(
     ],
     includes = ["include"],
     visibility = ["//visibility:public"],
-    alwayslink = True,
 )
 
 cc_library(

FILES.md

@@ -250,12 +250,16 @@ Further documentation:
 
 ### `BUILD.bazel`
 
-The file can be updated by calling
+The build definition for [Bazel](https://bazel.build). The file is generated by
+`cmake/scripts/gen_bazel_build_file.cmake`, which derives the header list from the files in `include`; change the
+script rather than editing the file by hand. The file can be updated by calling
 
 ```shell
 make BUILD.bazel
 ```
 
+The "Check amalgamation" workflow fails if the file is out of date.
+
 ### `meson.build`
 
 The build definition for the [Meson](https://mesonbuild.com) build system.

Makefile

@@ -1,4 +1,4 @@
-.PHONY: pretty clean ChangeLog.md release update_hedley update_hedley_undef
+.PHONY: pretty clean ChangeLog.md release update_hedley update_hedley_undef BUILD.bazel
 
 ##########################################################################
 # configuration
@@ -30,8 +30,9 @@ AMALGAMATED_FWD_FILE=single_include/nlohmann/json_fwd.hpp
 # main target
 all:
 	@echo "amalgamate - amalgamate files single_include/nlohmann/json{,_fwd}.hpp from the include/nlohmann sources"
+	@echo "BUILD.bazel - regenerate the Bazel BUILD file from the include/nlohmann sources"
 	@echo "ChangeLog.md - generate ChangeLog file"
-	@echo "check-amalgamation - check whether sources have been amalgamated"
+	@echo "check-amalgamation - check whether sources have been amalgamated and BUILD.bazel is up to date"
 	@echo "clean - remove built files"
 	@echo "doctest - compile example files and check their output"
 	@echo "fuzz_testing - prepare fuzz testing of the JSON parser"
@@ -172,8 +173,13 @@ check-amalgamation:
 	@diff $(AMALGAMATED_FWD_FILE) $(AMALGAMATED_FWD_FILE)~ || (echo "===================================================================\n  Amalgamation required! Please read the contribution guidelines\n  in file .github/CONTRIBUTING.md.\n===================================================================" ; mv $(AMALGAMATED_FWD_FILE)~ $(AMALGAMATED_FWD_FILE) ; false)
 	@mv $(AMALGAMATED_FILE)~ $(AMALGAMATED_FILE)
 	@mv $(AMALGAMATED_FWD_FILE)~ $(AMALGAMATED_FWD_FILE)
+	@mv BUILD.bazel BUILD.bazel~
+	@$(MAKE) BUILD.bazel
+	@diff BUILD.bazel BUILD.bazel~ || (echo "===================================================================\n  BUILD.bazel is out of date! Please run 'make BUILD.bazel'.\n===================================================================" ; mv BUILD.bazel~ BUILD.bazel ; false)
+	@mv BUILD.bazel~ BUILD.bazel
 
-BUILD.bazel: $(SRCS)
+# generate the Bazel BUILD file; phony, because a removed header would not trigger a rebuild
+BUILD.bazel:
 	cmake -P cmake/scripts/gen_bazel_build_file.cmake
 
 ##########################################################################

cmake/ci.cmake

@@ -230,21 +230,6 @@ add_custom_target(ci_test_simdutf
     COMMENT "Compile and test with simdutf UTF-8 validation enabled"
 )
 
-###############################################################################
-# Enable brace-init copy semantics.
-###############################################################################
-
-add_custom_target(ci_test_brace_init_copy_semantics
-    COMMAND ${CMAKE_COMMAND}
-    -DCMAKE_BUILD_TYPE=Debug -GNinja
-    -DJSON_BuildTests=ON -DJSON_FastTests=ON
-    -DCMAKE_CXX_FLAGS=-DJSON_BRACE_INIT_COPY_SEMANTICS=1
-    -S${PROJECT_SOURCE_DIR} -B${PROJECT_BINARY_DIR}/build_brace_init_copy_semantics
-    COMMAND ${CMAKE_COMMAND} --build ${PROJECT_BINARY_DIR}/build_brace_init_copy_semantics
-    COMMAND cd ${PROJECT_BINARY_DIR}/build_brace_init_copy_semantics && ${CMAKE_CTEST_COMMAND} --parallel ${N} --output-on-failure
-    COMMENT "Compile and test with brace-init copy semantics enabled"
-)
-
 ###############################################################################
 # Enable strict NUL-byte handling.
 ###############################################################################
@@ -311,6 +296,25 @@ add_custom_target(ci_test_skiplibraryversioncheck
     COMMENT "Compile and run a translation unit simulating a mismatched library version, with JSON_SKIP_LIBRARY_VERSION_CHECK defined"
 )
 
+###############################################################################
+# Disable thread-local storage.
+###############################################################################
+
+# Without thread-local storage, copying and comparing cannot bound their
+# descent and handle every object and array without the call stack. Those paths
+# are otherwise only reached by values nested deeper than the bound, so this
+# target is what runs the whole test suite through them.
+add_custom_target(ci_test_no_thread_local
+    COMMAND ${CMAKE_COMMAND}
+    -DCMAKE_BUILD_TYPE=Debug -GNinja
+    -DJSON_BuildTests=ON
+    -DCMAKE_CXX_FLAGS=-DJSON_NO_THREAD_LOCAL
+    -S${PROJECT_SOURCE_DIR} -B${PROJECT_BINARY_DIR}/build_no_thread_local
+    COMMAND ${CMAKE_COMMAND} --build ${PROJECT_BINARY_DIR}/build_no_thread_local
+    COMMAND cd ${PROJECT_BINARY_DIR}/build_no_thread_local && ${CMAKE_CTEST_COMMAND} --parallel ${N} --output-on-failure
+    COMMENT "Compile and test without thread-local storage"
+)
+
 ###############################################################################
 # Coverage.
 ###############################################################################
@@ -615,7 +619,7 @@ add_custom_target(ci_single_binaries
 add_custom_target(ci_benchmarks
     COMMAND ${CMAKE_COMMAND}
         -DCMAKE_BUILD_TYPE=Release -GNinja
-        -S${PROJECT_SOURCE_DIR}/tests/benchmarks -B${PROJECT_BINARY_DIR}/build_benchmarks
+        -S${PROJECT_SOURCE_DIR}/benchmarks -B${PROJECT_BINARY_DIR}/build_benchmarks
     COMMAND ${CMAKE_COMMAND} --build ${PROJECT_BINARY_DIR}/build_benchmarks --target json_benchmarks
     COMMAND cd ${PROJECT_BINARY_DIR}/build_benchmarks && ./json_benchmarks
     COMMENT "Run benchmarks"

cmake/download_test_data.cmake

@@ -77,7 +77,7 @@ if(CMAKE_CROSSCOMPILING)
 endif()
 if(NOT DEFINED LIBCPP_VERSION_OUTPUT_CACHED)
     try_run(RUN_RESULT_VAR COMPILE_RESULT_VAR
-        "${CMAKE_BINARY_DIR}" SOURCES "${CMAKE_CURRENT_LIST_DIR}/detect_libcpp_version.cpp"
+        "${CMAKE_BINARY_DIR}" SOURCES "${CMAKE_SOURCE_DIR}/cmake/detect_libcpp_version.cpp"
         RUN_OUTPUT_VARIABLE LIBCPP_VERSION_OUTPUT
         COMPILE_OUTPUT_VARIABLE LIBCPP_VERSION_COMPILE_OUTPUT
     )

cmake/scripts/gen_bazel_build_file.cmake

@@ -1,24 +1,57 @@
 # generate Bazel BUILD file
+#
+# usage: cmake -P cmake/scripts/gen_bazel_build_file.cmake (or: make BUILD.bazel)
+#
+# The header list of the "json" target is derived from the files in include/. Everything else is fixed text below,
+# so edit this script rather than BUILD.bazel.
 
-set(PROJECT_ROOT "${CMAKE_CURRENT_LIST_DIR}/../..")
+get_filename_component(PROJECT_ROOT "${CMAKE_CURRENT_LIST_DIR}/../.." ABSOLUTE)
 set(BUILD_FILE "${PROJECT_ROOT}/BUILD.bazel")
 
-file(GLOB_RECURSE HEADERS LIST_DIRECTORIES false RELATIVE "${PROJECT_ROOT}" "include/*.hpp")
+file(GLOB_RECURSE HEADERS LIST_DIRECTORIES false RELATIVE "${PROJECT_ROOT}" "${PROJECT_ROOT}/include/*.hpp")
+list(SORT HEADERS)
+
+set(CONTENT [=[
+load("@rules_cc//cc:cc_library.bzl", "cc_library")
+load("@rules_license//rules:license.bzl", "license")
+
+package(
+    default_applicable_licenses = [":license"],
+)
+
+exports_files([
+    "LICENSE.MIT",
+])
+
+license(
+    name = "license",
+    license_kinds = ["@rules_license//licenses/spdx:MIT"],
+    license_text = "LICENSE.MIT",
+)
 
-file(WRITE "${BUILD_FILE}" [=[
 cc_library(
     name = "json",
     hdrs = [
 ]=])
 
 foreach(header ${HEADERS})
-    file(APPEND "${BUILD_FILE}" "        \"${header}\",\n")
+    string(APPEND CONTENT "        \"${header}\",\n")
 endforeach()
 
-file(APPEND "${BUILD_FILE}" [=[
+string(APPEND CONTENT [=[
     ],
     includes = ["include"],
     visibility = ["//visibility:public"],
-    alwayslink = True,
+)
+
+cc_library(
+    name = "singleheader-json",
+    hdrs = [
+        "single_include/nlohmann/json.hpp",
+    ],
+    includes = ["single_include"],
+    visibility = ["//visibility:public"],
 )
 ]=])
+
+file(WRITE "${BUILD_FILE}" "${CONTENT}")

docs/mkdocs/docs/api/basic_json/get_to.md

@@ -21,6 +21,10 @@ This overload is chosen if:
 - `ValueType` is not `basic_json`,
 - `json_serializer<ValueType>` has a `from_json()` method of the form `void from_json(const basic_json&, ValueType&)`
 
+`v` must not be `const`. Passing a `const` object is a compile-time error. For types such as arithmetic types, enums,
+and C arrays, the error is a `static_assert` that names the problem. For other types, the overload is not viable, and
+the compiler reports that no matching `get_to` was found.
+
 ## Template parameters
 
 `ValueType`
@@ -67,3 +71,4 @@ Depends on the `json_serializer<ValueType>::from_json()` implementation.
 ## Version history
 
 - Since version 3.3.0.
+- Added a `static_assert` with a clear message for `const` arguments in version 3.13.0.

docs/mkdocs/docs/api/basic_json/sax_parse.md

@@ -69,7 +69,9 @@ The SAX event lister must follow the interface of [`json_sax`](../json_sax/index
      [`input_format_t`](input_format_t.md) for more information
 
 `strict` (in)
-:   whether the input has to be consumed completely (optional, `#!cpp true` by default)
+:   whether the input has to be consumed completely (optional, `#!cpp true` by default); when `#!cpp false` and the
+    input is a `#!cpp std::istream`, the character that terminates a number is consumed unless
+    [`JSON_PRECISE_STREAM_POSITION`](../macros/json_precise_stream_position.md) is defined to `1`; see [`operator>>`](../operator_gtgt.md#notes)
 
 `ignore_comments` (in)
 :   whether comments should be ignored and treated like whitespace (`#!cpp true`) or yield a parse error
@@ -136,6 +138,8 @@ A UTF-8 byte order mark is silently ignored.
 - Added `ignore_trailing_commas` in version 3.13.0.
 - Extended container support (1) to include types with lvalue-only ADL `begin`/`end` (matching `std::begin`/`std::end` semantics) in version 3.13.0.
 - Extended overload (2) to accept heterogeneous iterator+sentinel pairs (C++20 ranges support) in version 3.13.0.
+- `JSON_PRECISE_STREAM_POSITION` added in version 3.13.0 to optionally leave a `#!cpp std::istream` positioned right
+  after the parsed value when `strict` is `#!cpp false`.
 
 !!! warning "Deprecation"
 

docs/mkdocs/docs/api/macros/index.md

@@ -16,6 +16,8 @@ header. See also the [macro overview page](../../features/macros.md).
 
 ## Parsing
 
+- [**JSON_PRECISE_STREAM_POSITION**](json_precise_stream_position.md) - opt in to leaving an input stream positioned
+  right after a parsed number
 - [**JSON_STRICT_NUL_HANDLING**](json_strict_nul_handling.md) - opt in to rejecting a NUL byte in the input instead of
   treating it as end of input
 
@@ -27,6 +29,7 @@ header. See also the [macro overview page](../../features/macros.md).
 - [**JSON_HAS_STD_FORMAT**](json_has_std_format.md) - control `std::format`/`std::formatter` support
 - [**JSON_HAS_THREE_WAY_COMPARISON**](json_has_three_way_comparison.md) - control 3-way comparison support
 - [**JSON_NO_IO**](json_no_io.md) - switch off functions relying on certain C++ I/O headers
+- [**JSON_NO_THREAD_LOCAL**](json_no_thread_local.md) - switch off the use of `thread_local` storage
 - [**JSON_SKIP_UNSUPPORTED_COMPILER_CHECK**](json_skip_unsupported_compiler_check.md) - do not warn about unsupported compilers
 - [**JSON_USE_GLOBAL_UDLS**](json_use_global_udls.md) - place user-defined string literals (UDLs) into the global namespace
 - [**JSON_USE_SIMDUTF**](json_use_simdutf.md) - use the simdutf library to accelerate UTF-8 validation

docs/mkdocs/docs/api/macros/json_brace_init_copy_semantics.md

@@ -38,6 +38,28 @@ The default value is `0` (disabled — existing behavior is preserved).
 
     This macro must be defined **before** including `<nlohmann/json.hpp>`. Defining it after the include has no effect.
 
+!!! warning "Applies to every single-element list"
+
+    The macro does not only affect a single JSON value in braces. **Any** single-element braced list is treated as its
+    element, so it no longer creates a one-element array:
+
+    ```cpp
+    json j1 = {1};       // 1, not [1]
+    json j2 = {"text"};  // "text", not ["text"]
+    json j3 = {{1, 2}};  // [1,2], not [[1,2]]
+    ```
+
+    Code that relies on these producing arrays must use `json::array()` instead (see below). Lists with more than one
+    element, and a single `[string, value]` pair such as `{{"key", "value"}}`, which still creates an object, are not
+    affected. The library's own conversions are not affected either: for example, `std::tuple<int>{5}` still becomes
+    `[5]`.
+
+!!! note "ABI compatibility"
+
+    The value of this macro is encoded in the [namespace](../../features/namespace.md) (tag `_bics`), resulting in
+    distinct symbol names. Translation units compiled with and without it can therefore be linked into the same program
+    without One Definition Rule (ODR) violations, but they cannot exchange instances of library types.
+
 !!! tip "Workaround without the macro"
 
     To explicitly create a single-element array without enabling this macro, use `json::array()`:

docs/mkdocs/docs/api/macros/json_no_thread_local.md

@@ -0,0 +1,48 @@
+# JSON_NO_THREAD_LOCAL
+
+```cpp
+#define JSON_NO_THREAD_LOCAL
+```
+
+When defined, the library does not use `#!cpp thread_local` storage. This is relevant for the few environments whose
+toolchain does not support it.
+
+Copying a value and comparing two values both descend into the first levels by letting the containers copy or compare
+themselves, and finish whatever is nested deeper than that without the call stack, so that neither can exhaust the stack
+however deeply the values are nested. Each counts the levels it has descended into in a `#!cpp thread_local` variable, as
+a counter shared between threads would be raced.
+
+Without those counters, no descent can be bounded safely, so objects and arrays are copied and compared without the call
+stack right away. Both keep working exactly as they do otherwise - the same values come out, the same comparisons hold,
+and deeply nested values are handled just as safely - but both are slower, because the containers no longer copy or
+compare themselves. Copying the benchmark documents takes 9% (`canada.json`) to 34% (`twitter.json`) longer, and
+comparing two equal ones 10% (`citm_catalog.json`) to 90% (`canada.json`) longer.
+
+## Default definition
+
+By default, `#!cpp JSON_NO_THREAD_LOCAL` is not defined.
+
+```cpp
+#undef JSON_NO_THREAD_LOCAL
+```
+
+The library defines it by itself for Clang targeting MinGW, which does not survive the `#!cpp thread_local` storage:
+copying a value segfaults there, with both old and current Clang versions, while GCC targeting MinGW is unaffected.
+Copying and comparing fall back to working without the call stack there, as they do whenever the macro is defined.
+
+## Examples
+
+??? example
+
+    The code below forces the library not to use `#!cpp thread_local` storage.
+
+    ```cpp
+    #define JSON_NO_THREAD_LOCAL 1
+    #include <nlohmann/json.hpp>
+
+    ...
+    ```
+
+## Version history
+
+- Added in version 3.12.1.

docs/mkdocs/docs/api/macros/json_precise_stream_position.md

@@ -0,0 +1,131 @@
+# JSON_PRECISE_STREAM_POSITION
+
+```cpp
+#define JSON_PRECISE_STREAM_POSITION /* value */
+```
+
+When defined to `1`, [`operator>>`](../operator_gtgt.md) and [`sax_parse`](../basic_json/sax_parse.md) with
+`strict = false` leave a `#!cpp std::istream` positioned right after the parsed value for every value type. By default,
+the character that terminates a number is consumed as well.
+
+The macro only affects reading from a `#!cpp std::istream` when the rest of the stream is not required to be consumed.
+[`parse`](../basic_json/parse.md), [`accept`](../basic_json/accept.md), and all other inputs (strings, iterators,
+containers, `#!cpp FILE*`) are never affected.
+
+## Default definition
+
+The default value is `0` (disabled — existing behavior is preserved).
+
+```cpp
+#define JSON_PRECISE_STREAM_POSITION 0
+```
+
+## Notes
+
+!!! note "Background"
+
+    A number is the only JSON value whose end can be detected solely by reading the character that follows it. By
+    default, that character is consumed and not put back, so the stream is left one byte too far after a number, and
+    only after a number:
+
+    ```cpp
+    std::istringstream input("1true");
+    json j;
+    input >> j;  // j == 1, but the stream now starts at "rue"
+    ```
+
+    With this macro, the character is only looked at and left in the stream, so the stream starts at `true`. This
+    does not require the stream buffer to support putting a character back.
+
+    This was not changed unconditionally, because code can depend on the consumed character, even unknowingly (see
+    [#5340](https://github.com/nlohmann/json/issues/5340)). Both of the following work by default only because the
+    character after each number is swallowed, and behave differently with this macro:
+
+    ```cpp
+    std::istringstream input("1,2,3");
+    json j1, j2, j3;
+    input >> j1 >> j2 >> j3;  // default: 1, 2, 3
+                              // with the macro: throws parse_error.101 at the ','
+    ```
+
+    ```cpp
+    std::istringstream input("42\nfoo");
+    json j;
+    std::string line;
+    input >> j;
+    std::getline(input, line);  // default: "foo"
+                                // with the macro: "" (like after reading an int with >>)
+    ```
+
+    In both cases, the behavior with the macro is what you already get today when the value is not a number: `"a","b"`
+    fails at the `,`, and `std::getline` after `{}` returns an empty string. This macro offers an opt-in path to
+    the consistent behavior ahead of version 4.0.0, where it is planned to become the default.
+
+!!! warning "Opt-in only"
+
+    This macro must be defined **before** including `<nlohmann/json.hpp>`. Defining it after the include has no
+    effect.
+
+!!! note "ABI compatibility"
+
+    The value of this macro is encoded in the [namespace](../../features/namespace.md) (tag `_psp`), resulting in
+    distinct symbol names. Translation units compiled with and without it can therefore be linked into the same program
+    without One Definition Rule (ODR) violations, but they cannot exchange instances of library types.
+
+!!! tip "Workaround without the macro"
+
+    Separate the values in the stream with whitespace. The character consumed after a number is then the separator,
+    and whitespace before the next value is skipped anyway.
+
+## Examples
+
+??? example "Default behavior (macro not defined)"
+
+    Without the macro, the character after a number is consumed:
+
+    ```cpp
+    #include <iostream>
+    #include <sstream>
+    #include <nlohmann/json.hpp>
+
+    using json = nlohmann::json;
+
+    int main()
+    {
+        std::istringstream input("1true");
+        json j1, j2;
+        input >> j1;  // j1 == 1
+        input >> j2;  // throws parse_error.101: the stream now starts at "rue"
+    }
+    ```
+
+??? example "Opt-in precise stream position (macro defined to 1)"
+
+    With the macro, the stream is positioned right after the number:
+
+    ```cpp
+    #define JSON_PRECISE_STREAM_POSITION 1
+    #include <iostream>
+    #include <sstream>
+    #include <nlohmann/json.hpp>
+
+    using json = nlohmann::json;
+
+    int main()
+    {
+        std::istringstream input("1true");
+        json j1, j2;
+        input >> j1;  // j1 == 1
+        input >> j2;  // j2 == true
+    }
+    ```
+
+## See also
+
+- [**operator>>**](../operator_gtgt.md) - deserialize from stream
+- [**sax_parse**](../basic_json/sax_parse.md) - generate SAX events
+
+## Version history
+
+- Added in version 3.13.0.
+- Planned to become the default (with the macro removed) in version 4.0.0.

docs/mkdocs/docs/api/macros/json_use_implicit_conversions.md

@@ -24,6 +24,14 @@ By default, implicit conversions are enabled.
     You can prepare existing code by already defining `JSON_USE_IMPLICIT_CONVERSIONS` to `0` and replace any implicit
     conversions with calls to [`get`](../basic_json/get.md).
 
+!!! tip "Automatic migration"
+
+    The community-maintained clang-tidy check `modernize-nlohmann-json-explicit-conversions` rewrites implicit
+    conversions into explicit calls to [`get`](../basic_json/get.md); for example, `#!cpp int i = j;` becomes
+    `#!cpp int i = j.get<int>();`. The check is not part of clang-tidy itself, and it does not catch every case (for
+    example, constructing a `std::optional` from a JSON value), so review the result. See
+    [discussion #4610](https://github.com/nlohmann/json/discussions/4610) for how to build and use it.
+
 !!! hint "CMake option"
 
     Implicit conversions can also be controlled with the CMake option

docs/mkdocs/docs/api/operator_gtgt.md

@@ -67,7 +67,9 @@ input >> j2;  // parses the next value
     Only numbers are affected. Values ending in a self-delimiting character do not read past themselves, so
     `truefalse`, `[1][2]`, `{"a":1}{"b":2}`, and `"a""b"` can be read back to back without a separator.
 
-    This is tracked in [#5340](https://github.com/nlohmann/json/issues/5340).
+    Define [`JSON_PRECISE_STREAM_POSITION`](macros/json_precise_stream_position.md) to `1` to leave the terminating character in the stream
+    instead, so that the stream is positioned right after the value for every value type and no separator is
+    needed. This is tracked in [#5340](https://github.com/nlohmann/json/issues/5340).
 
 Note that reading concatenated values does **not** work for [JSON Lines](../features/parsing/json_lines.md)
 (newline-delimited JSON) input -- see that page for why and for the recommended alternative.
@@ -107,9 +109,12 @@ being read.
 - [parse](basic_json/parse.md) - deserialize from a compatible input
 - [`JSON_STRICT_NUL_HANDLING`](macros/json_strict_nul_handling.md) - opt in to rejecting a NUL byte in the input
   instead of treating it as end of input
+- [`JSON_PRECISE_STREAM_POSITION`](macros/json_precise_stream_position.md) - opt in to leaving the stream positioned right after a number
 
 ## Version history
 
 - Added in version 1.0.0.
 - `JSON_STRICT_NUL_HANDLING` added in version 3.13.0 to optionally reject a NUL byte in the input instead of treating
   it as end of input; planned to become the default in version 4.0.0.
+- `JSON_PRECISE_STREAM_POSITION` added in version 3.13.0 to optionally leave the character that terminates a number in
+  the stream; planned to become the default in version 4.0.0.

docs/mkdocs/docs/community/assurance_case.md

@@ -0,0 +1,70 @@
+# Assurance case
+
+This page argues why the library meets its security requirements. It describes the threats the library faces, where the
+trust boundaries lie, and how the library's design and the [quality assurance](quality_assurance.md) counter these
+threats. To report a vulnerability, see the [security policy](security_policy.md).
+
+## Threat model
+
+The library parses, stores, and serializes JSON values in memory. It does not open network connections, does not open
+files (it only reads from streams or `std::FILE*` handles that the caller has already opened), does not read environment
+variables, and does not implement cryptography or handle credentials.
+
+The primary threat is therefore **untrusted input**: JSON text or binary data (BJData, BSON, CBOR, MessagePack, UBJSON)
+that an attacker controls, passed to [`parse`](../api/basic_json/parse.md), [`accept`](../api/basic_json/accept.md),
+[`sax_parse`](../api/basic_json/sax_parse.md), or one of the `from_*` functions such as
+[`from_cbor`](../api/basic_json/from_cbor.md). Such input may try to
+
+- make the library read or write out of bounds (malformed lengths, truncated input, invalid UTF-8),
+- trigger undefined behavior (integer overflow in sizes or numbers, invalid casts),
+- exhaust memory (huge announced sizes), or
+- exhaust the call stack (deeply nested arrays and objects).
+
+## Trust boundaries
+
+- **Untrusted:** all serialized input read by the parser, the SAX interface, and the binary readers. The library must
+  handle every possible input by either producing a value or throwing a [`parse_error`](../home/exceptions.md#parse-errors)
+  (or returning `false` when exceptions are disabled for the call).
+- **Trusted:** the C++ code that calls the library. Calling a function with violated preconditions, for instance
+  accessing an array with [`operator[]`](../api/basic_json/operator%5B%5D.md) out of range, is a programming error and
+  not a security boundary. Such preconditions are checked with [runtime assertions](../features/assertions.md) in debug
+  builds; functions such as [`at`](../api/basic_json/at.md) offer checked access with exceptions.
+
+## Secure design
+
+- **Strict parsing.** The parser accepts exactly the JSON grammar of [RFC 8259](https://datatracker.ietf.org/doc/html/rfc8259).
+  Extensions such as [comments](../features/comments.md) and [trailing commas](../features/trailing_commas.md) must be
+  enabled explicitly. Invalid UTF-8 is rejected.
+- **Errors are reported, not ignored.** Malformed input results in a [`parse_error`](../home/exceptions.md#parse-errors)
+  with the byte position of the error. Binary readers do not trust announced sizes: strings and binary values grow
+  only as bytes are actually read, arrays reserve at most a fixed number of elements up front, and sizes that no
+  container can hold are rejected.
+- **Memory is owned by values.** Each `basic_json` value owns its content, and there is no manual memory management in
+  user code. The destructor does not recurse, so destroying a deeply nested value does not exhaust the stack.
+- **Bounded recursion.** The JSON parser and the binary readers keep their state in explicit stacks instead of
+  recursing per nesting level. Operations that walk a value, such as [`dump`](../api/basic_json/dump.md), copying,
+  hashing, and [`merge_patch`](../api/basic_json/merge_patch.md), recurse only up to a fixed depth and continue with an
+  explicit stack below it. Some operations, such as comparison, [`diff`](../api/basic_json/diff.md),
+  [`flatten`](../api/basic_json/flatten.md), and the binary writers, still recurse once per nesting level; work on them
+  is in progress. Applications that process untrusted input can limit its nesting depth with a
+  [parser callback](../features/parsing/parser_callbacks.md).
+- **Invariants are checked.** The class invariant (for instance, that the pointer for the stored type is never null) is
+  checked with runtime assertions throughout the test suite.
+
+## Common weaknesses
+
+The following table maps the relevant classes of the [Common Weakness Enumeration](https://cwe.mitre.org) to the
+measures that counter them. The measures are described in detail in [Quality assurance](quality_assurance.md).
+
+| Weakness                                                                  | Countermeasures                                                                                  |
+|---------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------|
+| Out-of-bounds read/write ([CWE-125](https://cwe.mitre.org/data/definitions/125.html), [CWE-787](https://cwe.mitre.org/data/definitions/787.html)) | bounds checks on all reads from the input; AddressSanitizer and Valgrind on the test suite; OSS-Fuzz      |
+| Integer overflow ([CWE-190](https://cwe.mitre.org/data/definitions/190.html)) | UndefinedBehaviorSanitizer with integer overflow detection; Clang-Tidy; Cppcheck                 |
+| Use after free, double free ([CWE-416](https://cwe.mitre.org/data/definitions/416.html), [CWE-415](https://cwe.mitre.org/data/definitions/415.html)) | ownership of all memory by values; AddressSanitizer and Valgrind; Clang Static Analyzer          |
+| Memory leaks ([CWE-401](https://cwe.mitre.org/data/definitions/401.html)) | Valgrind (Memcheck) on the test suite                                                            |
+| Uncontrolled recursion ([CWE-674](https://cwe.mitre.org/data/definitions/674.html)) | iterative parser, binary readers, and destructor; bounded recursion in value operations; tests with deeply nested inputs |
+| Uncontrolled resource consumption ([CWE-400](https://cwe.mitre.org/data/definitions/400.html)) | allocations based on announced sizes are capped; OSS-Fuzz with memory limits                     |
+| Undefined behavior in general ([CWE-758](https://cwe.mitre.org/data/definitions/758.html)) | UndefinedBehaviorSanitizer; runtime assertions; Clang-Tidy, Cppcheck, Clang Static Analyzer, Infer |
+
+In addition, every line of the library is covered by the unit tests, and all parsers are fuzz-tested around the clock
+by [OSS-Fuzz](https://github.com/google/oss-fuzz/tree/master/projects/json).

docs/mkdocs/docs/community/index.md

@@ -5,4 +5,6 @@
 - [Contribution Guidelines](contribution_guidelines.md) - guidelines how to contribute to this project
 - [Governance](governance.md) - the governance model of this project
 - [Quality Assurance](quality_assurance.md) - how the quality of this project is assured
+- [Roadmap](roadmap.md) - what the project will and will not do
 - [Security Policy](security_policy.md) - the security policy of the project
+- [Assurance Case](assurance_case.md) - why the library meets its security requirements

docs/mkdocs/docs/community/quality_assurance.md

@@ -164,6 +164,9 @@ Note: Some modern features (like C++20 ranges or filesystem support) may be disa
 - [x] The parser is tested against extensive correctness suites for JSON compliance.
 - [x] In addition, the library is continuously fuzz-tested at [OSS-Fuzz](https://google.github.io/oss-fuzz/) where the
   library is checked against billions of inputs.
+- [x] Every crash reported by OSS-Fuzz is fixed together with a unit test that reproduces it, and the fix references
+  the OSS-Fuzz issue. The round-trip checks of the fuzzer drivers are also part of the unit tests. See the
+  [fuzz testing documentation](https://github.com/nlohmann/json/blob/develop/tests/fuzzing.md#handling-oss-fuzz-reports).
 
 ## Static analysis
 

docs/mkdocs/docs/community/roadmap.md

@@ -0,0 +1,43 @@
+# Roadmap
+
+This page describes what the project intends to do, and what it does not intend to do, over the next year. Concrete
+work items are tracked in the [GitHub milestones](https://github.com/nlohmann/json/milestones) and the
+[issue tracker](https://github.com/nlohmann/json/issues).
+
+## What the project will do
+
+- **Keep the C++11 baseline.** The library will continue to compile with every
+  [supported C++11 compiler](https://github.com/nlohmann/json/blob/develop/README.md#supported-compilers). Features of
+  later standards are only used when they are guarded by the `JSON_HAS_CPP_*` macros.
+- **Stay conformant to JSON.** The parser and serializer follow [RFC 8259](https://datatracker.ietf.org/doc/html/rfc8259).
+  Extensions such as [comments](../features/comments.md) or [trailing commas](../features/trailing_commas.md) remain
+  opt-in.
+- **Keep the 3.x public API stable.** Releases follow [semantic versioning](https://semver.org). Changes that would
+  break existing code are only added behind a feature macro, so users can opt in and test their code before a next
+  major release.
+- **Support a broad range of compilers and platforms.** The [CI](quality_assurance.md) keeps testing old and new
+  versions of GCC, Clang, MSVC, and other compilers on Linux, macOS, and Windows.
+- **Keep the quality assurance up.** Every change keeps the test coverage at 100%, passes the static and dynamic
+  analysis, and is fuzz-tested by OSS-Fuzz, see [Quality assurance](quality_assurance.md).
+- **Harden the library against hostile input.** Handling deeply nested values without exhausting the call stack is
+  ongoing work.
+- **Fix bugs and security issues** reported through the issue tracker and the [security policy](security_policy.md).
+
+## What the project will not do
+
+- **Break the public API of version 3.x.** See the
+  [contribution guidelines](https://github.com/nlohmann/json/blob/develop/.github/CONTRIBUTING.md#break-the-public-api)
+  for what counts as a breaking change.
+- **Require a newer C++ standard than C++11.**
+- **Break JSON conformance** or enable non-standard extensions by default.
+- **Add dependencies** or require a build step. The library remains header-only, and the single header
+  `json.hpp` remains a complete distribution.
+- **Trade simplicity for speed or memory efficiency.** Performance improvements are welcome, but the library is not
+  meant to compete with the fastest JSON libraries, see [Design goals](../home/design_goals.md).
+
+## Version 4.0
+
+There is no decision yet on whether or when a version 4.0 with breaking changes will be released. Proposals that need
+a major version, for instance stricter type conversions, are collected in issue
+[#3453](https://github.com/nlohmann/json/issues/3453). Until then, such changes are only added as opt-in behavior
+behind feature macros.

docs/mkdocs/docs/features/binary_formats/bjdata.md

@@ -116,18 +116,22 @@ The library uses the following mapping from JSON values types to BJData types ac
     ```
 
     Likewise, when a JSON object in the above form is serialized using
-    [`to_bjdata`](../../api/basic_json/to_bjdata.md), it is automatically converted into a compact BJData ND-array. When
-    the 1-dimensional vector stored in `"_ArraySize_"` contains a single integer or two integers with one being 1, a
-    regular 1-D optimized array is generated instead.
+    [`to_bjdata`](../../api/basic_json/to_bjdata.md), it is automatically converted into a compact BJData ND-array.
 
-    An object is only converted if the annotation actually describes a packed array; otherwise it is serialized as a
-    regular JSON object. This requires all of the following:
+    When parsing, an ND-array whose dimension vector is empty, contains a single integer, contains two integers with the
+    first being 1, or contains a 0 is returned as a regular (possibly empty) array rather than an annotated object.
+
+    An object is only converted if the annotation describes a packed array that is parsed back into the same annotated
+    object; otherwise it is serialized as a regular JSON object, so the annotation is never lost in a round trip. This requires
+    all of the following:
 
     - `"_ArrayType_"` is one of `uint8`, `int8`, `uint16`, `int16`, `uint32`, `int32`, `uint64`, `int64`, `single`,
       `double`, `char`, or `byte`,
     - `"_ArraySize_"` is an array, since the dimensions are written as the ND-array header's length,
-    - every entry of `"_ArraySize_"` is a non-negative integer, and their product is representable as a `std::size_t`,
-    - `"_ArrayData_"` holds exactly that many elements, and
+    - `"_ArraySize_"` has at least two entries and is not a 1×N row vector (first entry 1), since other shapes are
+      parsed back as a regular array,
+    - every entry of `"_ArraySize_"` is a positive integer, and their product is representable as a `std::size_t`,
+    - `"_ArrayData_"` is an array holding exactly that many elements, and
     - every element of `"_ArrayData_"` is a number of the kind named by `"_ArrayType_"` (a floating-point number for
       `single` and `double`, an integer otherwise).
 
@@ -204,6 +208,16 @@ The library maps BJData types to JSON value types as follows:
 
     The mapping is **complete** in the sense that any BJData value can be converted to a JSON value.
 
+!!! info "Round trips"
+
+    A value returned by [`from_bjdata`](../../api/basic_json/from_bjdata.md) can be serialized with
+    [`to_bjdata`](../../api/basic_json/to_bjdata.md) using any combination of options and parsed back into an equal
+    value, and serializing that value again with the same options produces the same bytes. The exception is binary
+    values: they are only written as an optimized binary array (`[$B`) if Draft 3 is enabled and both `use_size` and
+    `use_type` are set. Otherwise, they are written as arrays of integers and parsed back as such (see the notes on
+    binary values above), and serializing such an array again may choose different, but equally valid, type markers.
+    The bytes can then differ, but parsing them again yields the same value.
+
 ??? example
 
     ```cpp

docs/mkdocs/docs/features/macros.md

@@ -91,6 +91,23 @@ security reasons (e.g., Intel Software Guard Extensions (SGX)).
 
 See [full documentation of `JSON_NO_IO`](../api/macros/json_no_io.md).
 
+## `JSON_NO_THREAD_LOCAL`
+
+When defined, the library does not use `#!cpp thread_local` storage. Copying a value and comparing two values then
+always avoid the call stack rather than descending into a bounded number of levels first, which is slower but yields the
+same values and the same comparisons.
+
+See [full documentation of `JSON_NO_THREAD_LOCAL`](../api/macros/json_no_thread_local.md).
+
+## `JSON_PRECISE_STREAM_POSITION`
+
+When defined to `1`, [`operator>>`](../api/operator_gtgt.md) and non-strict
+[`sax_parse`](../api/basic_json/sax_parse.md) leave an input stream positioned right after the parsed value, instead of
+also consuming the character that terminates a number. The default value is `0`, which preserves the existing behavior;
+this is planned to become the default in version 4.0.0.
+
+See [full documentation of `JSON_PRECISE_STREAM_POSITION`](../api/macros/json_precise_stream_position.md).
+
 ## `JSON_SKIP_LIBRARY_VERSION_CHECK`
 
 When defined, the library will not create a compiler warning when a different version of the library was already

docs/mkdocs/docs/features/namespace.md

@@ -15,6 +15,10 @@ The complete default namespace name is derived as follows:
     - [`JSON_DIAGNOSTICS`](../api/macros/json_diagnostics.md) defined non-zero appends `_diag`.
     - [`JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON`](../api/macros/json_use_legacy_discarded_value_comparison.md)
       defined non-zero appends `_ldvcmp`.
+    - [`JSON_DIAGNOSTIC_POSITIONS`](../api/macros/json_diagnostic_positions.md) defined non-zero appends `_dp`.
+    - [`JSON_BRACE_INIT_COPY_SEMANTICS`](../api/macros/json_brace_init_copy_semantics.md) defined non-zero appends
+      `_bics`.
+    - [`JSON_PRECISE_STREAM_POSITION`](../api/macros/json_precise_stream_position.md) defined non-zero appends `_psp`.
 - The inline namespace ends with the suffix `_v` followed by the 3 components of the version number separated by
   underscores. To omit the version component, see [Disabling the version component](#disabling-the-version-component)
   below.

docs/mkdocs/docs/features/parsing/index.md

@@ -41,7 +41,8 @@ document followed by trailing bytes" is accepted rather than rejected. If you ar
 reject any input that is not exactly one JSON document, prefer `parse`.
 
 When using `operator>>` to read several concatenated values this way, a value that is a number must be followed by
-whitespace, because `operator>>` consumes the character that terminates a number — see the
+whitespace, because `operator>>` consumes the character that terminates a number, unless
+[`JSON_PRECISE_STREAM_POSITION`](../../api/macros/json_precise_stream_position.md) is defined to `1` — see the
 [`operator>>` notes](../../api/operator_gtgt.md#notes) for details and examples.
 
 ## SAX vs. DOM parsing

docs/mkdocs/docs/home/architecture.md

@@ -1,34 +1,125 @@
 # Architecture
 
-!!! info
-
-    This page is still under construction. Its goal is to provide a high-level overview of the library's architecture.
-    This should help new contributors to get an idea of the used concepts and where to make changes.
+This page gives a high-level overview of the library's architecture. It should help new contributors to get an idea of
+the used concepts and where to make changes.
 
 ## Overview
 
-The main structure is class [nlohmann::basic_json](../api/basic_json/index.md).
+The library is built around a single class template, [`nlohmann::basic_json`](../api/basic_json/index.md). A
+`basic_json` value is a node in a tree of JSON values. All other components either create such a tree from an input
+(parsing), write a tree to an output (serialization), or give access to it (iterators, JSON Pointer, conversions).
 
-- public API
-- container interface
-- iterators
+```mermaid
+flowchart LR
+    input[/"input<br>(string, stream,<br>iterator range, file)"/]
+    ia["input adapter"]
+    lexer["lexer"]
+    parser["parser"]
+    breader["binary_reader"]
+    sax["SAX interface"]
+    value[("basic_json<br>value tree")]
+    serializer["serializer"]
+    bwriter["binary_writer"]
+    oa["output adapter"]
+    output[/"output<br>(string, stream,<br>vector)"/]
 
-## Template specializations
+    input --> ia
+    ia --> lexer --> parser --> sax
+    ia --> breader --> sax
+    sax --> value
+    value --> serializer --> oa
+    value --> bwriter --> oa
+    oa --> output
+```
 
-- describe template parameters of `basic_json`
-- [`json`](../api/json.md)
-- [`ordered_json`](../api/ordered_json.md) via [`ordered_map`](../api/ordered_map.md)
+- **JSON text** is read by an [input adapter](#input-adapters), tokenized by the lexer, and turned into SAX events by
+  the parser.
+- **Binary formats** (BJData, BSON, CBOR, MessagePack, UBJSON) are read by an input adapter and turned into the same SAX
+  events by the `binary_reader`.
+- A [SAX consumer](#sax-interface) receives the events. The one used by [`parse`](../api/basic_json/parse.md) builds a
+  `basic_json` value tree.
+- The `serializer` (JSON text) or the `binary_writer` (binary formats) writes a value tree to an
+  [output adapter](#output-adapters).
+
+## Source layout
+
+The public headers are in [`include/nlohmann`](https://github.com/nlohmann/json/tree/develop/include/nlohmann):
+
+- [`json.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/json.hpp) defines class [`basic_json`](../api/basic_json/index.md).
+- [`json_fwd.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/json_fwd.hpp) contains forward declarations.
+- [`adl_serializer.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/adl_serializer.hpp), [`byte_container_with_subtype.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/byte_container_with_subtype.hpp), and [`ordered_map.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/ordered_map.hpp) define
+  [`adl_serializer`](../api/adl_serializer/index.md),
+  [`byte_container_with_subtype`](../api/byte_container_with_subtype/index.md), and
+  [`ordered_map`](../api/ordered_map.md).
+
+Everything else lives in [`detail/`](https://github.com/nlohmann/json/tree/develop/include/nlohmann/detail) and namespace `nlohmann::detail`, which is not part of the public API. Paths
+below are relative to `include/nlohmann`.
+
+| Component | Location |
+|-----------|----------|
+| Value type enumeration | [`detail/value_t.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/value_t.hpp) |
+| Input adapters | [`detail/input/input_adapters.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/input/input_adapters.hpp) |
+| Lexer | [`detail/input/lexer.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/input/lexer.hpp), [`detail/input/number_parse.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/input/number_parse.hpp), [`detail/input/string_scan.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/input/string_scan.hpp) |
+| Parser | [`detail/input/parser.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/input/parser.hpp) |
+| SAX interface and DOM builders | [`detail/input/json_sax.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/input/json_sax.hpp) |
+| Binary format readers | [`detail/input/binary_reader.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/input/binary_reader.hpp) |
+| JSON serializer | [`detail/output/serializer.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/output/serializer.hpp), [`detail/conversions/to_chars.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/conversions/to_chars.hpp) |
+| Binary format writers | [`detail/output/binary_writer.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/output/binary_writer.hpp) |
+| Output adapters | [`detail/output/output_adapters.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/output/output_adapters.hpp) |
+| Iterators | [`detail/iterators/`](https://github.com/nlohmann/json/tree/develop/include/nlohmann/detail/iterators) |
+| Conversions from/to arbitrary types | [`detail/conversions/from_json.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/conversions/from_json.hpp), [`detail/conversions/to_json.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/conversions/to_json.hpp) |
+| JSON Pointer | [`detail/json_pointer.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/json_pointer.hpp) |
+| Exceptions | [`detail/exceptions.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/exceptions.hpp) |
+| Type traits and C++ feature backports | [`detail/meta/`](https://github.com/nlohmann/json/tree/develop/include/nlohmann/detail/meta) |
+| Macros | [`detail/macro_scope.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/macro_scope.hpp), [`detail/macro_unscope.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/macro_unscope.hpp), [`detail/abi_macros.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/abi_macros.hpp) |
+
+The single-header version [`single_include/nlohmann/json.hpp`](https://github.com/nlohmann/json/blob/develop/single_include/nlohmann/json.hpp)
+is generated from these files with `make amalgamate` and must not be edited by hand.
+
+## Template parameters
+
+[`basic_json`](../api/basic_json/index.md) is parameterized by the types it uses to store values and to convert from and to other types:
+
+| Template parameter   | Default                     | Used for                                                          |
+|----------------------|-----------------------------|-------------------------------------------------------------------|
+| `ObjectType`         | `std::map`                  | objects, see [`object_t`](../api/basic_json/object_t.md)          |
+| `ArrayType`          | `std::vector`               | arrays, see [`array_t`](../api/basic_json/array_t.md)             |
+| `StringType`         | `std::string`               | strings and object keys, see [`string_t`](../api/basic_json/string_t.md) |
+| `BooleanType`        | `bool`                      | Booleans, see [`boolean_t`](../api/basic_json/boolean_t.md)       |
+| `NumberIntegerType`  | `std::int64_t`              | signed integers, see [`number_integer_t`](../api/basic_json/number_integer_t.md) |
+| `NumberUnsignedType` | `std::uint64_t`             | unsigned integers, see [`number_unsigned_t`](../api/basic_json/number_unsigned_t.md) |
+| `NumberFloatType`    | `double`                    | floating-point numbers, see [`number_float_t`](../api/basic_json/number_float_t.md) |
+| `AllocatorType`      | `std::allocator`            | allocating objects, arrays, strings, and binary values            |
+| `JSONSerializer`     | `adl_serializer`            | conversions from/to other types, see [`adl_serializer`](../api/adl_serializer/index.md) |
+| `BinaryType`         | `std::vector<std::uint8_t>` | binary values, see [`binary_t`](../api/basic_json/binary_t.md)  |
+| `CustomBaseClass`    | `void`                      | an optional base class, see [`json_base_class_t`](../api/basic_json/json_base_class_t.md) |
+
+The library provides two specializations:
+
+- [`json`](../api/json.md) uses all default template arguments.
+- [`ordered_json`](../api/ordered_json.md) uses [`ordered_map`](../api/ordered_map.md) as `ObjectType` to keep the
+  insertion order of object keys.
+
+The requirements on the template arguments are listed in
+[Template Parameter Requirements](../features/types/template_parameters.md).
 
 ## Value storage
 
-Values are stored as a tagged union of [value_t](../api/basic_json/value_t.md) and json_value.
+Each [`basic_json`](../api/basic_json/index.md) value stores its content as a tagged union: an enumeration [`value_t`](../api/basic_json/value_t.md)
+names the type of the value, and a union `json_value` holds the value itself. Both are members of the nested struct
+`data`, which is the only data member `m_data` of `basic_json`:
 
 ```cpp
-/// the type of the current element
-value_t m_type = value_t::null;
+struct data
+{
+    /// the type of the current element
+    value_t m_type = value_t::null;
 
-/// the value of the current element
-json_value m_value = {};
+    /// the value of the current element
+    json_value m_value = {};
+};
+
+data m_data = {};
 ```
 
 with
@@ -68,42 +159,83 @@ union json_value {
 };
 ```
 
-## Parsing inputs (deserialization)
+Objects, arrays, strings, and binary values are allocated on the heap with `AllocatorType`, and the union only stores a
+pointer to them. This keeps a `basic_json` value small: one pointer-sized union and one byte for the type. The class
+maintains the invariant that the pointer matching `m_type` is never null; `assert_invariant()` checks it with
+[runtime assertions](../features/assertions.md).
 
-Input is read via **input adapters** that abstract a source with a common interface:
+## Input adapters
+
+Input is read via **input adapters** that abstract a source. Every input adapter provides this interface:
 
 ```cpp
-/// read a single character
-std::char_traits<char>::int_type get_character() noexcept;
+/// the type of the characters in the input
+using char_type = ...;
 
-/// read multiple characters to a destination buffer and
-/// returns the number of characters successfully read
+/// read a single character; returns std::char_traits<char_type>::eof() at the end of the input
+typename std::char_traits<char_type>::int_type get_character();
+
+/// read up to count * sizeof(T) bytes into dest and return the number of bytes read
+/// (used by the binary readers)
 template<class T>
 std::size_t get_elements(T* dest, std::size_t count = 1);
 ```
 
-List examples of input adapters.
+The lexer detects two optional extensions at compile time. Only `iterator_input_adapter` provides them, and only for
+random-access input of single-byte characters:
 
-## SAX Interface
+- `supports_seek`, `get_consumed_count()`, and `copy_consumed_range()` let the lexer reconstruct already consumed input
+  for error messages instead of copying every character it reads.
+- `supports_bulk_scan`, `bulk_data()`, `bulk_remaining()`, and `bulk_skip()` let the lexer scan strings directly in
+  contiguous memory, several bytes at a time.
 
-TODO
+The function `input_adapter` picks the right adapter for the argument passed to `parse`, `accept`, `sax_parse`, or the
+`from_*` functions:
 
-## Writing outputs (serialization)
+- `iterator_input_adapter` reads from an iterator range, which also covers strings, containers, and pointers.
+- `wide_string_input_adapter` reads from ranges of `wchar_t`, `char16_t`, or `char32_t` and converts them to UTF-8.
+  It cannot be used for binary formats; its `get_elements()` throws.
+- `input_stream_adapter` reads from a `std::istream`.
+- `file_input_adapter` reads from a `std::FILE*`.
+
+## SAX interface
+
+The parser does not build values itself. It reports what it reads as events to a [SAX](../features/parsing/sax_interface.md)
+consumer, which implements the interface [`json_sax`](../api/json_sax/index.md): `null`, `boolean`, `number_integer`,
+`number_unsigned`, `number_float`, `string`, `binary`, `start_object`, `key`, `end_object`, `start_array`, `end_array`,
+and `parse_error`.
+
+The library comes with two consumers in `detail/input/json_sax.hpp`:
+
+- `json_sax_dom_parser` builds a [`basic_json`](../api/basic_json/index.md) value tree. [`parse`](../api/basic_json/parse.md) uses it.
+- `json_sax_dom_callback_parser` does the same, but calls a [parser callback](../features/parsing/parser_callbacks.md)
+  for each event, which can skip values. `parse` uses it when a callback is given.
+
+The `binary_reader` emits the same events for binary formats, so [`sax_parse`](../api/basic_json/sax_parse.md) works
+with a user-defined consumer for JSON and for all binary formats alike.
+
+## Output adapters
 
 Output is written via **output adapters**:
 
 ```cpp
-template<typename T>
 void write_character(CharType c);
 
-template<typename CharType>
 void write_characters(const CharType* s, std::size_t length);
 ```
 
-List examples of output adapters.
+The `serializer` (used by [`dump`](../api/basic_json/dump.md) and [`operator<<`](../api/operator_ltlt.md)) and the
+`binary_writer` (used by the `to_*` functions) write to one of these adapters:
+
+- `output_vector_adapter` appends to a `std::vector`.
+- `output_stream_adapter` writes to a `std::ostream`.
+- `output_string_adapter` appends to a string.
 
 ## Value conversion
 
+Values are converted from and to other types with the `JSONSerializer` template parameter. The default,
+[`adl_serializer`](../api/adl_serializer/index.md), calls the free functions
+
 ```cpp
 template<class T>
 void to_json(basic_json& j, const T& t);
@@ -112,13 +244,23 @@ template<class T>
 void from_json(const basic_json& j, T& t);
 ```
 
+found by argument-dependent lookup. The library defines them for standard types in `detail/conversions`; users add them
+for their own types, see [Arbitrary Type Conversions](../features/arbitrary_types.md). The
+[serialization macros](../features/macros.md) generate these functions.
+
 ## Additional features
 
-- JSON Pointers
-- Binary formats
-- Custom base class
-- Conversion macros
+- [JSON Pointer](../features/json_pointer.md) (class `json_pointer`) addresses values inside a tree. It is also the
+  basis of [JSON Patch](../features/json_patch.md).
+- [Binary formats](../features/binary_formats/index.md) are read by `binary_reader` and written by `binary_writer`.
+- A [custom base class](../api/basic_json/json_base_class_t.md) can add members to every [`basic_json`](../api/basic_json/index.md) value.
+- [Serialization macros](../features/macros.md) generate `to_json` and `from_json` functions for user-defined types.
 
 ## Details namespace
 
-- C++ feature backports
+Namespace `nlohmann::detail` contains all implementation details. It is not part of the public API and may change in any
+release. Besides the components above, it contains:
+
+- type traits to detect the capabilities of user-defined types (`detail/meta/type_traits.hpp`),
+- backports of C++14/17 features to C++11 (`detail/meta/cpp_future.hpp`), and
+- helpers such as `string_concat` and `string_escape`.

docs/mkdocs/docs/integration/migration_guide.md

@@ -176,6 +176,12 @@ You can prepare existing code by already defining
 conversions with calls to [`get`](../api/basic_json/get.md), [`get_to`](../api/basic_json/get_to.md),
 [`get_ref`](../api/basic_json/get_ref.md), or [`get_ptr`](../api/basic_json/get_ptr.md).
 
+!!! tip "Automatic migration"
+
+    The community-maintained clang-tidy check `modernize-nlohmann-json-explicit-conversions` rewrites most implicit
+    conversions into calls to [`get`](../api/basic_json/get.md). It is not part of clang-tidy itself; see
+    [discussion #4610](https://github.com/nlohmann/json/discussions/4610) for how to build and use it.
+
 === "Deprecated"
 
       ```cpp

docs/mkdocs/mkdocs.yml

@@ -292,6 +292,8 @@ nav:
           - 'JSON_HAS_THREE_WAY_COMPARISON': api/macros/json_has_three_way_comparison.md
           - 'JSON_NOEXCEPTION': api/macros/json_noexception.md
           - 'JSON_NO_IO': api/macros/json_no_io.md
+          - 'JSON_NO_THREAD_LOCAL': api/macros/json_no_thread_local.md
+          - 'JSON_PRECISE_STREAM_POSITION': api/macros/json_precise_stream_position.md
           - 'JSON_SKIP_LIBRARY_VERSION_CHECK': api/macros/json_skip_library_version_check.md
           - 'JSON_SKIP_UNSUPPORTED_COMPILER_CHECK': api/macros/json_skip_unsupported_compiler_check.md
           - 'JSON_STRICT_NUL_HANDLING': api/macros/json_strict_nul_handling.md
@@ -316,7 +318,9 @@ nav:
       - community/contribution_guidelines.md
       - community/quality_assurance.md
       - community/governance.md
+      - community/roadmap.md
       - community/security_policy.md
+      - community/assurance_case.md
 
 # Extras
 extra:

include/nlohmann/detail/abi_macros.hpp

@@ -34,6 +34,14 @@
     #define JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON 0
 #endif
 
+#ifndef JSON_BRACE_INIT_COPY_SEMANTICS
+    #define JSON_BRACE_INIT_COPY_SEMANTICS 0
+#endif
+
+#ifndef JSON_PRECISE_STREAM_POSITION
+    #define JSON_PRECISE_STREAM_POSITION 0
+#endif
+
 #if JSON_DIAGNOSTICS
     #define NLOHMANN_JSON_ABI_TAG_DIAGNOSTICS _diag
 #else
@@ -52,20 +60,34 @@
     #define NLOHMANN_JSON_ABI_TAG_LEGACY_DISCARDED_VALUE_COMPARISON
 #endif
 
+#if JSON_BRACE_INIT_COPY_SEMANTICS
+    #define NLOHMANN_JSON_ABI_TAG_BRACE_INIT_COPY_SEMANTICS _bics
+#else
+    #define NLOHMANN_JSON_ABI_TAG_BRACE_INIT_COPY_SEMANTICS
+#endif
+
+#if JSON_PRECISE_STREAM_POSITION
+    #define NLOHMANN_JSON_ABI_TAG_PRECISE_STREAM_POSITION _psp
+#else
+    #define NLOHMANN_JSON_ABI_TAG_PRECISE_STREAM_POSITION
+#endif
+
 #ifndef NLOHMANN_JSON_NAMESPACE_NO_VERSION
     #define NLOHMANN_JSON_NAMESPACE_NO_VERSION 0
 #endif
 
 // Construct the namespace ABI tags component
-#define NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c) json_abi ## a ## b ## c
-#define NLOHMANN_JSON_ABI_TAGS_CONCAT(a, b, c) \
-    NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c)
+#define NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e) json_abi ## a ## b ## c ## d ## e
+#define NLOHMANN_JSON_ABI_TAGS_CONCAT(a, b, c, d, e) \
+    NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e)
 
 #define NLOHMANN_JSON_ABI_TAGS                                       \
     NLOHMANN_JSON_ABI_TAGS_CONCAT(                                   \
             NLOHMANN_JSON_ABI_TAG_DIAGNOSTICS,                       \
             NLOHMANN_JSON_ABI_TAG_LEGACY_DISCARDED_VALUE_COMPARISON, \
-            NLOHMANN_JSON_ABI_TAG_DIAGNOSTIC_POSITIONS)
+            NLOHMANN_JSON_ABI_TAG_DIAGNOSTIC_POSITIONS,              \
+            NLOHMANN_JSON_ABI_TAG_BRACE_INIT_COPY_SEMANTICS,         \
+            NLOHMANN_JSON_ABI_TAG_PRECISE_STREAM_POSITION)
 
 // Construct the namespace version component
 #define NLOHMANN_JSON_NAMESPACE_VERSION_CONCAT_EX(major, minor, patch) \

include/nlohmann/detail/conversions/to_json.hpp

@@ -471,6 +471,30 @@ inline void to_json_tuple_impl(BasicJsonType& j, const Tuple& t, index_sequence<
     j = { std::get<Idx>(t)... };
 }
 
+#if JSON_BRACE_INIT_COPY_SEMANTICS
+// JSON_BRACE_INIT_COPY_SEMANTICS makes a one-element braced list copy its
+// element instead of wrapping it, which would serialize std::tuple<int>{5} as 5
+// rather than [5]. Build what the default deduction builds instead: an object
+// if the element is a [string, value] pair, a one-element array otherwise.
+template<typename BasicJsonType, typename Tuple>
+inline void to_json_tuple_impl(BasicJsonType& j, const Tuple& t, index_sequence<0> /*unused*/)
+{
+    BasicJsonType element(std::get<0>(t));
+    // same test as the initializer-list constructor, including the cast that
+    // keeps a string type constructible from 0 from selecting operator[](key)
+    const bool is_member = element.is_array() && element.size() == 2
+                           && element[static_cast<typename BasicJsonType::size_type>(0)].is_string();
+    if (is_member)
+    {
+        j = BasicJsonType::object({std::move(element)});
+    }
+    else
+    {
+        j = BasicJsonType::array({std::move(element)});
+    }
+}
+#endif
+
 template<typename BasicJsonType, typename Tuple>
 inline void to_json_tuple_impl(BasicJsonType& j, const Tuple& /*unused*/, index_sequence<> /*unused*/)
 {

include/nlohmann/detail/hash.hpp

@@ -11,8 +11,10 @@
 #include <cstdint> // uint8_t
 #include <cstddef> // size_t
 #include <functional> // hash
+#include <vector> // vector
 
 #include <nlohmann/detail/abi_macros.hpp>
+#include <nlohmann/detail/recursion_depth_limit.hpp>
 #include <nlohmann/detail/value_t.hpp>
 
 NLOHMANN_JSON_NAMESPACE_BEGIN
@@ -26,6 +28,9 @@ inline std::size_t combine(std::size_t seed, std::size_t h) noexcept
     return seed;
 }
 
+template<typename BasicJsonType>
+std::size_t hash_iteratively(const BasicJsonType& j);
+
 /*!
 @brief hash a JSON value
 
@@ -33,12 +38,21 @@ The hash function tries to rely on std::hash where possible. Furthermore, the
 type of the JSON value is taken into account to have different hash values for
 null, 0, 0U, and false, etc.
 
+Hashing an array or an object hashes its elements, which used to call this
+function again once per nesting level, so a value nested deeply enough
+exhausted the call stack and terminated the process. The descent is bounded
+here: once @ref recursion_depth_limit levels have been entered, @ref
+hash_iteratively hashes what is left without the call stack. A value nested
+less deeply than that - all but a vanishing minority - is hashed exactly as
+before, without allocating.
+
 @tparam BasicJsonType basic_json specialization
 @param j JSON value to hash
+@param depth nesting level of @a j, counted from the value passed by the caller
 @return hash value of j
 */
 template<typename BasicJsonType>
-std::size_t hash(const BasicJsonType& j)
+std::size_t hash(const BasicJsonType& j, const std::size_t depth = 0)
 {
     using string_t = typename BasicJsonType::string_t;
     using number_integer_t = typename BasicJsonType::number_integer_t;
@@ -56,22 +70,32 @@ std::size_t hash(const BasicJsonType& j)
 
         case BasicJsonType::value_t::object:
         {
+            if (JSON_HEDLEY_UNLIKELY(depth >= recursion_depth_limit()))
+            {
+                return hash_iteratively(j);
+            }
+
             auto seed = combine(type, j.size());
             for (const auto& element : j.items())
             {
                 const auto h = std::hash<string_t> {}(element.key());
                 seed = combine(seed, h);
-                seed = combine(seed, hash(element.value()));
+                seed = combine(seed, hash(element.value(), depth + 1));
             }
             return seed;
         }
 
         case BasicJsonType::value_t::array:
         {
+            if (JSON_HEDLEY_UNLIKELY(depth >= recursion_depth_limit()))
+            {
+                return hash_iteratively(j);
+            }
+
             auto seed = combine(type, j.size());
             for (const auto& element : j)
             {
-                seed = combine(seed, hash(element));
+                seed = combine(seed, hash(element, depth + 1));
             }
             return seed;
         }
@@ -127,5 +151,77 @@ std::size_t hash(const BasicJsonType& j)
     }
 }
 
+/// an array or object whose elements @ref hash_iteratively is hashing
+template<typename BasicJsonType>
+struct hash_frame
+{
+    hash_frame(const BasicJsonType* value_, std::size_t seed_) noexcept
+        : value(value_), position(value_->cbegin()), seed(seed_)
+    {}
+
+    const BasicJsonType* value;
+    typename BasicJsonType::const_iterator position;
+    std::size_t seed;
+};
+
+/*!
+@brief hash the array or object @a j without the call stack
+
+Computes the same value as @ref hash, keeping the arrays and objects it has
+entered on an explicit stack instead of descending into them. Only reached for
+values nested deeper than @ref recursion_depth_limit.
+
+@tparam BasicJsonType basic_json specialization
+@param j array or object to hash
+@return hash value of j
+*/
+template<typename BasicJsonType>
+std::size_t hash_iteratively(const BasicJsonType& j)
+{
+    using string_t = typename BasicJsonType::string_t;
+
+    std::vector<hash_frame<BasicJsonType>> stack;
+    stack.emplace_back(&j, combine(static_cast<std::size_t>(j.type()), j.size()));
+
+    while (true)
+    {
+        // a copy, as entering an element below can reallocate the stack; the
+        // frame itself is only changed through stack.back()
+        const hash_frame<BasicJsonType> frame = stack.back();
+
+        if (frame.position == frame.value->cend())
+        {
+            // all elements are hashed: fold this value's hash into its parent's
+            // seed, exactly where the recursive version returns it
+            const std::size_t h = frame.seed;
+            stack.pop_back();
+            if (stack.empty())
+            {
+                return h;
+            }
+            stack.back().seed = combine(stack.back().seed, h);
+            continue;
+        }
+
+        if (frame.value->is_object())
+        {
+            stack.back().seed = combine(stack.back().seed, std::hash<string_t> {}(frame.position.key()));
+        }
+
+        // advance before entering the element, which pushes onto the stack
+        const BasicJsonType& element = *frame.position;
+        ++stack.back().position;
+
+        if (element.is_structured())
+        {
+            stack.emplace_back(&element, combine(static_cast<std::size_t>(element.type()), element.size()));
+        }
+        else
+        {
+            stack.back().seed = combine(stack.back().seed, hash(element));
+        }
+    }
+}
+
 }  // namespace detail
 NLOHMANN_JSON_NAMESPACE_END

include/nlohmann/detail/input/input_adapters.hpp

@@ -101,6 +101,11 @@ class input_stream_adapter
         // maintain ifstream flags, except eof
         if (is != nullptr)
         {
+#if JSON_PRECISE_STREAM_POSITION
+            // consume the character last returned by get_character() unless it
+            // was given back with release_lookahead()
+            commit_lookahead();
+#endif
             is->clear(is->rdstate() & std::ios::eofbit);
         }
     }
@@ -114,6 +119,58 @@ class input_stream_adapter
     input_stream_adapter& operator=(input_stream_adapter&) = delete;
     input_stream_adapter& operator=(input_stream_adapter&&) = delete;
 
+#if JSON_PRECISE_STREAM_POSITION
+    input_stream_adapter(input_stream_adapter&& rhs) noexcept
+        : is(rhs.is), sb(rhs.sb), lookahead(rhs.lookahead)
+    {
+        rhs.is = nullptr;
+        rhs.sb = nullptr;
+        rhs.lookahead = false;
+    }
+
+    // Whether the character last returned by get_character() can be given back
+    // to the input with release_lookahead().
+    static constexpr bool supports_lookahead = true;
+
+    // std::istream/std::streambuf use std::char_traits<char>::to_int_type, to
+    // ensure that std::char_traits<char>::eof() and the character 0xFF do not
+    // end up as the same value, e.g., 0xFFFFFFFF.
+    //
+    // The character is peeked rather than consumed: it is only stepped over
+    // once the next character is requested, or when the adapter is destroyed.
+    // Until then, release_lookahead() can leave it in the input.
+    std::char_traits<char>::int_type get_character()
+    {
+        if (lookahead)
+        {
+            // step over the character returned by the previous call
+            sb->sbumpc();
+        }
+
+        auto res = sb->sgetc();
+        // set eof manually, as we don't use the istream interface.
+        if (JSON_HEDLEY_UNLIKELY(res == std::char_traits<char>::eof()))
+        {
+            // there is nothing to step over next time
+            lookahead = false;
+            is->clear(is->rdstate() | std::ios::eofbit);
+        }
+        else
+        {
+            lookahead = true;
+        }
+        return res;
+    }
+
+    // Leave the character last returned by get_character() in the input, so
+    // that the next read from the stream - by this adapter or by the caller
+    // once parsing is done - sees it again. Unlike putting a consumed
+    // character back, this cannot fail.
+    void release_lookahead() noexcept
+    {
+        lookahead = false;
+    }
+#else
     input_stream_adapter(input_stream_adapter&& rhs) noexcept
         : is(rhs.is), sb(rhs.sb)
     {
@@ -124,6 +181,9 @@ class input_stream_adapter
     // std::istream/std::streambuf use std::char_traits<char>::to_int_type, to
     // ensure that std::char_traits<char>::eof() and the character 0xFF do not
     // end up as the same value, e.g., 0xFFFFFFFF.
+    //
+    // The character is consumed, so the character that terminates a number
+    // stays consumed after parsing; see JSON_PRECISE_STREAM_POSITION.
     std::char_traits<char>::int_type get_character()
     {
         auto res = sb->sbumpc();
@@ -134,10 +194,14 @@ class input_stream_adapter
         }
         return res;
     }
+#endif
 
     template<class T>
     std::size_t get_elements(T* dest, std::size_t count = 1)
     {
+#if JSON_PRECISE_STREAM_POSITION
+        commit_lookahead();
+#endif
         auto res = static_cast<std::size_t>(sb->sgetn(reinterpret_cast<char*>(dest), static_cast<std::streamsize>(count * sizeof(T))));
         if (JSON_HEDLEY_UNLIKELY(res < count * sizeof(T)))
         {
@@ -147,9 +211,27 @@ class input_stream_adapter
     }
 
   private:
+#if JSON_PRECISE_STREAM_POSITION
+    // Step over the character last returned by get_character(). The character
+    // has already been peeked successfully, so for every streambuf with a get
+    // area this is a pointer increment that cannot fail.
+    void commit_lookahead()
+    {
+        if (lookahead)
+        {
+            lookahead = false;
+            sb->sbumpc();
+        }
+    }
+#endif
+
     /// the associated input stream
     std::istream* is = nullptr;
     std::streambuf* sb = nullptr;
+#if JSON_PRECISE_STREAM_POSITION
+    /// whether get_character() peeked a character that is not consumed yet
+    bool lookahead = false;
+#endif
 };
 #endif  // JSON_NO_IO
 

include/nlohmann/detail/input/lexer.hpp

@@ -127,6 +127,25 @@ constexpr bool input_adapter_supports_seek(std::false_type /*detected*/)
     return false;
 }
 
+// Detect whether an input adapter reads with one character of lookahead that
+// can be left in the input (see input_stream_adapter::supports_lookahead,
+// which is only defined with JSON_PRECISE_STREAM_POSITION), detected like
+// supports_seek above.
+template<typename InputAdapterType>
+using detect_supports_lookahead = decltype(InputAdapterType::supports_lookahead);
+
+template<typename InputAdapterType>
+constexpr bool input_adapter_supports_lookahead(std::true_type /*detected*/)
+{
+    return InputAdapterType::supports_lookahead;
+}
+
+template<typename InputAdapterType>
+constexpr bool input_adapter_supports_lookahead(std::false_type /*detected*/)
+{
+    return false;
+}
+
 // Detect whether an input adapter exposes a contiguous byte block that the
 // lexer can scan directly (see iterator_input_adapter::supports_bulk_scan).
 // Adapters without the flag - file, stream, wide-string, user-defined - fall
@@ -167,6 +186,12 @@ class lexer : public lexer_base<BasicJsonType>
     static constexpr bool lazy_token_string =
         input_adapter_supports_seek<InputAdapterType>(is_detected<detect_supports_seek, InputAdapterType> {});
 
+    /// whether a simulated unget can be passed on to the input adapter, which
+    /// then leaves the character in the input; see
+    /// input_adapter_supports_lookahead
+    static constexpr bool can_release_lookahead =
+        input_adapter_supports_lookahead<InputAdapterType>(is_detected<detect_supports_lookahead, InputAdapterType> {});
+
     /// whether string scanning may bulk-consume runs of ordinary characters
     /// directly from a contiguous input buffer (SWAR fast path). This requires
     /// the token to be reconstructible lazily (lazy_token_string), so bypassing
@@ -1898,6 +1923,21 @@ scan_number_done:
         uncapture_char(std::integral_constant<bool, lazy_token_string> {});
     }
 
+    /// adapter without lookahead: nothing to do (see release_lookahead)
+    void release_lookahead_impl(std::false_type /*can_release*/) const noexcept {}
+
+    /// adapter with lookahead: leave the character in the input instead
+    void release_lookahead_impl(std::true_type /*can_release*/)
+    {
+        if (next_unget)
+        {
+            // the character is read from the input again rather than replayed
+            // from current, so the adapter must not step over it
+            next_unget = false;
+            ia.release_lookahead();
+        }
+    }
+
     /// seekable adapter: nothing was captured, so nothing to undo
     void uncapture_char(std::true_type /*lazy*/) const noexcept {}
 
@@ -1961,6 +2001,31 @@ scan_number_done:
         return position;
     }
 
+    /*!
+    @brief pass a pending simulated unget on to the input
+
+    unget() only rewinds the lexer's own bookkeeping, so the character that
+    terminated the last token (e.g. the character after a number) would still
+    be stepped over when the input adapter is done. Callers that hand the
+    input back to the user afterwards - operator>> and non-strict sax_parse -
+    call this once when scanning is done, so that the input is positioned
+    right after the value.
+
+    Adapters without lookahead (see input_adapter_supports_lookahead) are not
+    handed back to the user, so this is a no-op for them. Without
+    JSON_PRECISE_STREAM_POSITION, no adapter has lookahead, so this is always a
+    no-op and the terminating character stays consumed.
+
+    Scanning may continue after this call: @a next_unget is cleared, and the
+    character is read from the input again instead of being replayed from
+    @a current. A pending unget of EOF needs no special case, because reaching
+    EOF leaves no lookahead to release.
+    */
+    void release_lookahead()
+    {
+        release_lookahead_impl(std::integral_constant<bool, can_release_lookahead> {});
+    }
+
 #if JSON_DIAGNOSTIC_POSITIONS
     /// return the offset of the first character of the last read token; unlike
     /// the token's parsed value, this accounts for escape sequences

include/nlohmann/detail/input/parser.hpp

@@ -100,13 +100,22 @@ class parser
             json_sax_dom_callback_parser<BasicJsonType, InputAdapterType> sdp(result, callback, allow_exceptions, &m_lexer);
             sax_parse_internal(&sdp);
 
-            // in strict mode, input must be completely read
-            if (strict && (get_token() != token_type::end_of_input))
+            if (strict)
             {
-                sdp.parse_error(m_lexer.get_position(),
-                                m_lexer.get_token_string(),
-                                parse_error::create(101, m_lexer.get_position(),
-                                                    exception_message(token_type::end_of_input, "value"), nullptr));
+                // in strict mode, input must be completely read
+                if (get_token() != token_type::end_of_input)
+                {
+                    sdp.parse_error(m_lexer.get_position(),
+                                    m_lexer.get_token_string(),
+                                    parse_error::create(101, m_lexer.get_position(),
+                                                        exception_message(token_type::end_of_input, "value"), nullptr));
+                }
+            }
+            else
+            {
+                // the caller keeps using the input: position it right after
+                // the value by leaving the character that terminated it
+                m_lexer.release_lookahead();
             }
 
             // in case of an error, return a discarded value
@@ -128,12 +137,20 @@ class parser
             json_sax_dom_parser<BasicJsonType, InputAdapterType> sdp(result, allow_exceptions, &m_lexer);
             sax_parse_internal(&sdp);
 
-            // in strict mode, input must be completely read
-            if (strict && (get_token() != token_type::end_of_input))
+            if (strict)
             {
-                sdp.parse_error(m_lexer.get_position(),
-                                m_lexer.get_token_string(),
-                                parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input, "value"), nullptr));
+                // in strict mode, input must be completely read
+                if (get_token() != token_type::end_of_input)
+                {
+                    sdp.parse_error(m_lexer.get_position(),
+                                    m_lexer.get_token_string(),
+                                    parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input, "value"), nullptr));
+                }
+            }
+            else
+            {
+                // see above
+                m_lexer.release_lookahead();
             }
 
             // in case of an error, return a discarded value
@@ -166,12 +183,24 @@ class parser
         (void)detail::is_sax_static_asserts<SAX, BasicJsonType> {};
         const bool result = sax_parse_internal(sax);
 
-        // strict mode: next byte must be EOF
-        if (result && strict && (get_token() != token_type::end_of_input))
+        if (result)
         {
-            return sax->parse_error(m_lexer.get_position(),
-                                    m_lexer.get_token_string(),
-                                    parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input, "value"), nullptr));
+            if (strict)
+            {
+                // strict mode: next byte must be EOF
+                if (get_token() != token_type::end_of_input)
+                {
+                    return sax->parse_error(m_lexer.get_position(),
+                                            m_lexer.get_token_string(),
+                                            parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input, "value"), nullptr));
+                }
+            }
+            else
+            {
+                // the caller keeps using the input: position it right after
+                // the value by leaving the character that terminated it
+                m_lexer.release_lookahead();
+            }
         }
 
         return result;

include/nlohmann/detail/macro_scope.hpp

@@ -186,6 +186,15 @@
     #define JSON_NO_UNIQUE_ADDRESS
 #endif
 
+// Clang targeting MinGW does not survive the thread_local storage the copy
+// constructor uses to bound its descent: every test that copies a value
+// segfaults with clang 11.0.1 and clang 18.1.8, while the same tests pass with
+// GCC targeting MinGW and with every other toolchain the library is tested on.
+// Copying works the same way without the counter, only more slowly.
+#if !defined(JSON_NO_THREAD_LOCAL) && defined(__clang__) && defined(__MINGW32__)
+    #define JSON_NO_THREAD_LOCAL 1
+#endif
+
 // disable documentation warnings on clang
 #if defined(__clang__)
     #pragma clang diagnostic push
@@ -804,10 +813,6 @@ void templated_json_throw(ExceptionType exception)
     #define JSON_USE_GLOBAL_UDLS 1
 #endif
 
-#ifndef JSON_BRACE_INIT_COPY_SEMANTICS
-    #define JSON_BRACE_INIT_COPY_SEMANTICS 0
-#endif
-
 #ifndef JSON_STRICT_NUL_HANDLING
     #define JSON_STRICT_NUL_HANDLING 0
 #endif

include/nlohmann/detail/macro_unscope.hpp

@@ -26,7 +26,6 @@
 #undef JSON_NO_UNIQUE_ADDRESS
 #undef JSON_DISABLE_ENUM_SERIALIZATION
 #undef JSON_USE_GLOBAL_UDLS
-#undef JSON_BRACE_INIT_COPY_SEMANTICS
 #undef JSON_STRICT_NUL_HANDLING
 
 #ifndef JSON_TEST_KEEP_MACROS
@@ -45,6 +44,8 @@
     #undef JSON_HAS_STD_FORMAT
     #undef JSON_HAS_STATIC_RTTI
     #undef JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON
+    #undef JSON_BRACE_INIT_COPY_SEMANTICS
+    #undef JSON_PRECISE_STREAM_POSITION
 #endif
 
 #include <nlohmann/thirdparty/hedley/hedley_undef.hpp>

include/nlohmann/detail/output/binary_writer.hpp

@@ -881,8 +881,6 @@ class binary_writer
                         return ubjson_prefix(v, use_bjdata) == first_prefix;
                     });
 
-                    std::vector<CharType> bjdx = {'[', '{', 'S', 'H', 'T', 'F', 'N', 'Z'}; // excluded markers in bjdata optimized type
-
                     // an optimized array of a valueless type carries no payload, so a
                     // reader has nothing but the declared count to bound the allocation
                     // by and refuses an excessive one. Write the unoptimized form for
@@ -893,7 +891,7 @@ class binary_writer
                                                      && j.m_data.m_value.array->size() > detail::max_valueless_container_size;
 
                     if (same_prefix && !excessive_valueless
-                            && !(use_bjdata && std::find(bjdx.begin(), bjdx.end(), first_prefix) != bjdx.end()))
+                            && !(use_bjdata && is_bjdata_excluded_type_marker(first_prefix)))
                     {
                         prefix_required = false;
                         oa.write_character(to_char_type('$'));
@@ -997,9 +995,7 @@ class binary_writer
                         return ubjson_prefix(v, use_bjdata) == first_prefix;
                     });
 
-                    std::vector<CharType> bjdx = {'[', '{', 'S', 'H', 'T', 'F', 'N', 'Z'}; // excluded markers in bjdata optimized type
-
-                    if (same_prefix && !(use_bjdata && std::find(bjdx.begin(), bjdx.end(), first_prefix) != bjdx.end()))
+                    if (same_prefix && !(use_bjdata && is_bjdata_excluded_type_marker(first_prefix)))
                     {
                         prefix_required = false;
                         oa.write_character(to_char_type('$'));
@@ -1726,6 +1722,21 @@ class binary_writer
         }
     }
 
+    /*!
+    @brief whether BJData forbids @a marker as the type of an optimized array
+           or object
+
+    Containers, strings, high-precision numbers, booleans and null cannot be
+    declared as the single type of an optimized container in BJData; such a
+    container is written unoptimized. The reader rejects them with the same
+    list (binary_reader::bjd_optimized_type_markers).
+    */
+    static constexpr bool is_bjdata_excluded_type_marker(const CharType marker) noexcept
+    {
+        return marker == '[' || marker == '{' || marker == 'S' || marker == 'H'
+               || marker == 'T' || marker == 'F' || marker == 'N' || marker == 'Z';
+    }
+
     static constexpr CharType get_ubjson_float_prefix(float /*unused*/)
     {
         return 'd';  // float 32
@@ -1800,8 +1811,19 @@ class binary_writer
             return true;
         }
 
-        std::size_t len = (value.at(key).empty() ? 0 : 1);
-        for (const auto& el : value.at(key))
+        // the reader only restores an annotated object from an ND-array header
+        // with at least two dimensions: an empty dimension vector, a single
+        // dimension, or a 1xN row vector is read back as a plain array, which
+        // would silently drop the annotation, so such an object falls back to
+        // a plain object encoding instead
+        const auto& dims = value.at(key);
+        if (dims.size() < 2 || (dims.size() == 2 && dims.at(0).is_number_integer() && dims.at(0).template get<std::int64_t>() == 1))
+        {
+            return true;
+        }
+
+        std::size_t len = 1;
+        for (const auto& el : dims)
         {
             // a dimension is read as an unsigned value below, so anything that
             // is not a non-negative integer is rejected: a non-integer entry
@@ -1823,15 +1845,26 @@ class binary_writer
                 return true;
             }
             const auto dim_size = static_cast<std::size_t>(dim);
-            if (dim_size != 0 && len > (std::numeric_limits<std::size_t>::max)() / dim_size)
+
+            // the reader turns an ND-array with any zero dimension into an
+            // empty plain array, dropping the annotation, so keep the object
+            if (dim_size == 0)
+            {
+                return true;
+            }
+            if (len > (std::numeric_limits<std::size_t>::max)() / dim_size)
             {
                 return true;
             }
             len *= dim_size;
         }
 
+        // the elements are written from _ArrayData_ as a flat list, so it has
+        // to be an array: size() is 0 for null and 1 for any other scalar, and
+        // iterating an object visits its values, so any of these could match
+        // the dimensions by accident and be encoded as an unrelated ND-array
         key = "_ArrayData_";
-        if (value.at(key).size() != len)
+        if (!value.at(key).is_array() || value.at(key).size() != len)
         {
             return true;
         }

include/nlohmann/detail/output/serializer.hpp

@@ -30,6 +30,7 @@
 #include <nlohmann/detail/meta/cpp_future.hpp>
 #include <nlohmann/detail/output/binary_writer.hpp>
 #include <nlohmann/detail/output/output_adapters.hpp>
+#include <nlohmann/detail/recursion_depth_limit.hpp>
 #include <nlohmann/detail/string_concat.hpp>
 #include <nlohmann/detail/value_t.hpp>
 
@@ -133,7 +134,7 @@ class serializer
 
     Serializing a container descends into its elements, so a value nested deeply
     enough used to exhaust the call stack and terminate the process with no
-    exception to catch. The descent is bounded here: once @ref dump_depth_limit
+    exception to catch. The descent is bounded here: once @ref recursion_depth_limit
     levels have been entered, @ref dump_iteratively writes out what is left
     without the call stack. A value nested less deeply than that - all but a
     vanishing minority - is written by exactly the code that always wrote it.
@@ -148,7 +149,7 @@ class serializer
         {
             case value_t::object:
             {
-                if (JSON_HEDLEY_UNLIKELY(depth >= dump_depth_limit()))
+                if (JSON_HEDLEY_UNLIKELY(depth >= recursion_depth_limit()))
                 {
                     dump_iteratively(val, current_indent);
                     return;
@@ -223,7 +224,7 @@ class serializer
 
             case value_t::array:
             {
-                if (JSON_HEDLEY_UNLIKELY(depth >= dump_depth_limit()))
+                if (JSON_HEDLEY_UNLIKELY(depth >= recursion_depth_limit()))
                 {
                     dump_iteratively(val, current_indent);
                     return;
@@ -408,19 +409,12 @@ class serializer
     }
 
   private:
-    /// the number of levels @ref dump_internal descends into before it hands
-    /// over to @ref dump_iteratively
-    static constexpr std::size_t dump_depth_limit()
-    {
-        return 128;
-    }
-
     /*!
     @brief write out @a val and everything below it without the call stack
 
     Emits the same bytes as @ref dump_internal, keeping the containers it has
     entered on an explicit stack instead of descending into them. Only reached
-    for values nested deeper than @ref dump_depth_limit, which is why it is not
+    for values nested deeper than @ref recursion_depth_limit, which is why it is not
     written for speed: walking every value this way measured up to 20% slower on
     object-heavy documents than letting the compiler drive the descent.
     */

include/nlohmann/detail/recursion_depth_limit.hpp

@@ -0,0 +1,35 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  JSON for Modern C++
+// |  |  |__   |  |  | | | |  version 3.12.0
+// |_____|_____|_____|_|___|  https://github.com/nlohmann/json
+//
+// SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>
+// SPDX-License-Identifier: MIT
+
+#pragma once
+
+#include <cstddef> // size_t
+
+#include <nlohmann/detail/abi_macros.hpp>
+
+NLOHMANN_JSON_NAMESPACE_BEGIN
+namespace detail
+{
+
+/*!
+@brief the number of nesting levels an operation recurses into
+
+Operations that walk a value (serializing, hashing, merging, ...) recurse once
+per nesting level, which is fastest, but a value nested deeply enough would
+exhaust the call stack. So they recurse only this many levels deep and finish
+whatever lies below with an explicit stack. All of them share this limit.
+
+@sa https://github.com/nlohmann/json/issues/5387
+*/
+constexpr std::size_t recursion_depth_limit() noexcept
+{
+    return 128;
+}
+
+}  // namespace detail
+NLOHMANN_JSON_NAMESPACE_END

single_include/nlohmann/json_fwd.hpp

@@ -52,6 +52,14 @@
     #define JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON 0
 #endif
 
+#ifndef JSON_BRACE_INIT_COPY_SEMANTICS
+    #define JSON_BRACE_INIT_COPY_SEMANTICS 0
+#endif
+
+#ifndef JSON_PRECISE_STREAM_POSITION
+    #define JSON_PRECISE_STREAM_POSITION 0
+#endif
+
 #if JSON_DIAGNOSTICS
     #define NLOHMANN_JSON_ABI_TAG_DIAGNOSTICS _diag
 #else
@@ -70,20 +78,34 @@
     #define NLOHMANN_JSON_ABI_TAG_LEGACY_DISCARDED_VALUE_COMPARISON
 #endif
 
+#if JSON_BRACE_INIT_COPY_SEMANTICS
+    #define NLOHMANN_JSON_ABI_TAG_BRACE_INIT_COPY_SEMANTICS _bics
+#else
+    #define NLOHMANN_JSON_ABI_TAG_BRACE_INIT_COPY_SEMANTICS
+#endif
+
+#if JSON_PRECISE_STREAM_POSITION
+    #define NLOHMANN_JSON_ABI_TAG_PRECISE_STREAM_POSITION _psp
+#else
+    #define NLOHMANN_JSON_ABI_TAG_PRECISE_STREAM_POSITION
+#endif
+
 #ifndef NLOHMANN_JSON_NAMESPACE_NO_VERSION
     #define NLOHMANN_JSON_NAMESPACE_NO_VERSION 0
 #endif
 
 // Construct the namespace ABI tags component
-#define NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c) json_abi ## a ## b ## c
-#define NLOHMANN_JSON_ABI_TAGS_CONCAT(a, b, c) \
-    NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c)
+#define NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e) json_abi ## a ## b ## c ## d ## e
+#define NLOHMANN_JSON_ABI_TAGS_CONCAT(a, b, c, d, e) \
+    NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e)
 
 #define NLOHMANN_JSON_ABI_TAGS                                       \
     NLOHMANN_JSON_ABI_TAGS_CONCAT(                                   \
             NLOHMANN_JSON_ABI_TAG_DIAGNOSTICS,                       \
             NLOHMANN_JSON_ABI_TAG_LEGACY_DISCARDED_VALUE_COMPARISON, \
-            NLOHMANN_JSON_ABI_TAG_DIAGNOSTIC_POSITIONS)
+            NLOHMANN_JSON_ABI_TAG_DIAGNOSTIC_POSITIONS,              \
+            NLOHMANN_JSON_ABI_TAG_BRACE_INIT_COPY_SEMANTICS,         \
+            NLOHMANN_JSON_ABI_TAG_PRECISE_STREAM_POSITION)
 
 // Construct the namespace version component
 #define NLOHMANN_JSON_NAMESPACE_VERSION_CONCAT_EX(major, minor, patch) \

tests/CMakeLists.txt

@@ -75,7 +75,15 @@ target_compile_options(test_main PUBLIC
     #                              is annotated JSON_HEDLEY_NO_RETURN (it always throws), which
     #                              makes MSVC flag the code following its call in binary_reader.hpp
     #                              as unreachable for that instantiation, in both Debug and Release
-    $<$<CXX_COMPILER_ID:MSVC>:/W4;/wd4566;/wd4996;/wd4702>
+    #       Disable warning C4503: decorated name length exceeded, name was truncated; the deep
+    #                              copy support added for #5387 pushes the mangled name of
+    #                              std::allocator_traits<...>::construct for the custom-base-class
+    #                              test's map type past VS2015's limit. The name is only used for
+    #                              debug info, so truncation does not affect the build.
+    #       Disable warning C5285: cannot declare a specialization for 'std::tuple'; MSVC 19.51
+    #                              reports the forward declarations of standard library
+    #                              templates in the vendored doctest.h
+    $<$<CXX_COMPILER_ID:MSVC>:/W4;/wd4566;/wd4996;/wd4702;/wd4503;/wd5285>
     # https://github.com/nlohmann/json/issues/1114
     $<$<CXX_COMPILER_ID:MSVC>:/bigobj> $<$<BOOL:${MINGW}>:-Wa,-mbig-obj>
 

tests/abi/config/CMakeLists.txt

@@ -14,6 +14,20 @@ add_test(
     NAME test-abi_config_noversion
     COMMAND abi_config_noversion ${DOCTEST_TEST_FILTER})
 
+# test default and no version namespace with all ABI tags enabled, so the
+# expected tag order is checked regardless of the JSON_* CMake options
+foreach(test default noversion)
+    add_executable(abi_config_${test}_all_tags ${test}.cpp)
+    target_compile_definitions(abi_config_${test}_all_tags PRIVATE
+        JSON_DIAGNOSTICS=1
+        JSON_DIAGNOSTIC_POSITIONS=1
+        JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON=1)
+    target_link_libraries(abi_config_${test}_all_tags PRIVATE abi_compat_main)
+    add_test(
+        NAME test-abi_config_${test}_all_tags
+        COMMAND abi_config_${test}_all_tags ${DOCTEST_TEST_FILTER})
+endforeach()
+
 # test custom namespace
 add_executable(abi_config_custom custom.cpp)
 target_link_libraries(abi_config_custom PRIVATE abi_compat_main)

tests/abi/config/default.cpp

@@ -24,12 +24,20 @@ TEST_CASE("default namespace")
         expected += "_diag";
 #endif
 
+#if JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON
+        expected += "_ldvcmp";
+#endif
+
 #if JSON_DIAGNOSTIC_POSITIONS
         expected += "_dp";
 #endif
 
-#if JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON
-        expected += "_ldvcmp";
+#if JSON_BRACE_INIT_COPY_SEMANTICS
+        expected += "_bics";
+#endif
+
+#if JSON_PRECISE_STREAM_POSITION
+        expected += "_psp";
 #endif
 
         expected += "_v" STRINGIZE(NLOHMANN_JSON_VERSION_MAJOR);

tests/abi/config/noversion.cpp

@@ -25,12 +25,20 @@ TEST_CASE("default namespace without version component")
         expected += "_diag";
 #endif
 
+#if JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON
+        expected += "_ldvcmp";
+#endif
+
 #if JSON_DIAGNOSTIC_POSITIONS
         expected += "_dp";
 #endif
 
-#if JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON
-        expected += "_ldvcmp";
+#if JSON_BRACE_INIT_COPY_SEMANTICS
+        expected += "_bics";
+#endif
+
+#if JSON_PRECISE_STREAM_POSITION
+        expected += "_psp";
 #endif
 
         expected += "::basic_json";

tests/benchmarks/CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.14)
+cmake_minimum_required(VERSION 3.11...3.14)
 project(JSON_Benchmarks LANGUAGES CXX)
 
 # set compiler flags
@@ -6,35 +6,29 @@ if((CMAKE_CXX_COMPILER_ID MATCHES GNU) OR (CMAKE_CXX_COMPILER_ID MATCHES Clang))
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -flto -DNDEBUG -O3")
 endif()
 
-# configure Google Benchmark; a fixed release, so that results stay comparable
-set(JSON_GOOGLE_BENCHMARK_VERSION 1.9.5)
+# configure Google Benchmarks
 include(FetchContent)
-
-# only the library is needed; -Werror would break the pinned release as soon as
-# a newer compiler adds a warning
-set(BENCHMARK_ENABLE_TESTING OFF CACHE BOOL "" FORCE)
-set(BENCHMARK_ENABLE_INSTALL OFF CACHE BOOL "" FORCE)
-set(BENCHMARK_ENABLE_WERROR OFF CACHE BOOL "" FORCE)
-
-FetchContent_Declare(benchmark
-    URL https://github.com/google/benchmark/archive/refs/tags/v${JSON_GOOGLE_BENCHMARK_VERSION}.tar.gz
-    URL_HASH SHA256=9631341c82bac4a288bef951f8b26b41f69021794184ece969f8473977eaa340
-    DOWNLOAD_EXTRACT_TIMESTAMP TRUE
+FetchContent_Declare(
+    benchmark
+    GIT_REPOSITORY https://github.com/google/benchmark.git
+    GIT_TAG origin/main
+    GIT_SHALLOW TRUE
 )
-FetchContent_MakeAvailable(benchmark)
+
+FetchContent_GetProperties(benchmark)
+if(NOT benchmark_POPULATED)
+    FetchContent_Populate(benchmark)
+    set(BENCHMARK_ENABLE_TESTING OFF CACHE INTERNAL "" FORCE)
+    add_subdirectory(${benchmark_SOURCE_DIR} ${benchmark_BINARY_DIR})
+endif()
 
 # download test data
 set(CMAKE_MODULE_PATH ${CMAKE_CURRENT_SOURCE_DIR}/../../cmake ${CMAKE_MODULE_PATH})
 include(download_test_data)
 
-# the header to benchmark; point this at a directory holding another version's
-# nlohmann/json.hpp to compare versions (see README.md)
-set(JSON_BENCHMARK_INCLUDE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/../../single_include" CACHE PATH
-    "directory containing the nlohmann/json.hpp to benchmark")
-
 # benchmark binary
 add_executable(json_benchmarks src/benchmarks.cpp)
 target_compile_features(json_benchmarks PRIVATE cxx_std_11)
 target_link_libraries(json_benchmarks benchmark ${CMAKE_THREAD_LIBS_INIT})
 add_dependencies(json_benchmarks download_test_data)
-target_include_directories(json_benchmarks PRIVATE ${JSON_BENCHMARK_INCLUDE_DIR} ${CMAKE_BINARY_DIR}/include)
+target_include_directories(json_benchmarks PRIVATE ${CMAKE_SOURCE_DIR}/../../single_include ${CMAKE_BINARY_DIR}/include)

tests/benchmarks/README.md

@@ -1,130 +0,0 @@
-# Benchmarks
-
-Micro-benchmarks for parsing, serialization and the binary formats, written with
-[Google Benchmark](https://github.com/google/benchmark). They are not run by CI; see
-[When to run them](#when-to-run-them).
-
-## What is measured
-
-| benchmark | what it does |
-|---|---|
-| `ParseFile`, `ParseString` | parse JSON from a file stream or a string |
-| `ParseIndented` | parse the large files re-indented by 4 spaces, for the lexer's whitespace handling |
-| `Dump` | serialize, compact (`-`) and indented (`4`) |
-| `ToCbor`, `BinaryToCbor` | write CBOR; `BinaryToCbor` writes binary values of growing size |
-| `FromMsgpack` | read MessagePack; unchanged over the years, so its numbers stay comparable across releases |
-| `FromBinaryBuffer`, `FromBinaryFile` | read CBOR, MessagePack, UBJSON, BJData and BSON from a buffer or a `FILE*` |
-| `FromBinaryShape` | read deeply nested, container-heavy and scalar-heavy documents in every binary format |
-| `FromCborChunkedString` | read CBOR strings split into indefinite-length chunks |
-
-The input files are those of [nativejson-benchmark](https://github.com/miloyip/nativejson-benchmark) (`canada`,
-`citm_catalog`, `twitter`), a large `jeopardy` file, and number-heavy files (`floats`, `signed_ints`, ...).
-`bytes_per_second` counts the bytes read or written: the JSON text when parsing, the output when serializing.
-
-## Requirements
-
-- CMake 3.14 or later, a C++11 compiler, and Ninja for the `make` target.
-- Network access on the first configure: CMake downloads Google Benchmark and the
-  [test data](https://github.com/nlohmann/json_test_data) into the build directory. To reuse a download of the test
-  data, pass `-DJSON_TestDataDirectory=<build directory>/test_files`.
-- Google Benchmark is pinned to a release (1.9.5), so that results from different days stay comparable. To update it,
-  change `JSON_GOOGLE_BENCHMARK_VERSION` and the archive's `URL_HASH` in `CMakeLists.txt` together.
-- The benchmarks include `single_include/nlohmann/json.hpp`, so run `make amalgamate` after changing anything in
-  `include/`.
-
-GCC and Clang builds use `-O3 -flto -DNDEBUG`.
-
-## Running them
-
-From the repository root, this builds everything from scratch in `cmake-build-benchmarks` and runs all benchmarks:
-
-```sh
-make run_benchmarks
-```
-
-To build once and run selectively:
-
-```sh
-cmake -S tests/benchmarks -B build-benchmarks -G Ninja -DCMAKE_BUILD_TYPE=Release
-cmake --build build-benchmarks
-build-benchmarks/json_benchmarks --benchmark_filter='ParseString|Dump'
-```
-
-Useful options of `json_benchmarks`:
-
-| option | effect |
-|---|---|
-| `--benchmark_list_tests` | list the benchmarks instead of running them |
-| `--benchmark_filter=<regex>` | run only the benchmarks whose names match |
-| `--benchmark_repetitions=<n>` | run every benchmark `n` times and add mean, median, standard deviation and coefficient of variation |
-| `--benchmark_enable_random_interleaving=true` | run the repetitions in random order, which spreads out drifts such as thermal throttling |
-| `--benchmark_min_time=<seconds>s` | run each benchmark at least this long (e.g. `2s`) |
-| `--benchmark_out=<file> --benchmark_out_format=json` | also write the results to a file, e.g. for `compare.py` |
-
-## Reading the output
-
-Each line shows the wall-clock `Time` and the `CPU` time per iteration, the number of `Iterations` Google Benchmark
-chose, and the throughput in `bytes_per_second`. With repetitions, the lines ending in `_median` are the ones to
-compare. A `_cv` (coefficient of variation) above a few percent means the machine was too noisy for small
-differences to mean anything.
-
-## Comparing two versions
-
-To see what a change or a release did, build the same benchmarks twice: once against the header of the version to
-compare with, and once against the current one. `JSON_BENCHMARK_INCLUDE_DIR` names the directory holding the
-`nlohmann/json.hpp` to benchmark. For example, to compare the current checkout with 3.12.0:
-
-```sh
-# the header of the version to compare with
-mkdir -p build-baseline-header/nlohmann
-git show v3.12.0:single_include/nlohmann/json.hpp > build-baseline-header/nlohmann/json.hpp
-
-# the same benchmarks, built against either header
-cmake -S tests/benchmarks -B build-baseline -G Ninja -DCMAKE_BUILD_TYPE=Release \
-      -DJSON_BENCHMARK_INCLUDE_DIR="$PWD/build-baseline-header"
-cmake -S tests/benchmarks -B build-current -G Ninja -DCMAKE_BUILD_TYPE=Release
-cmake --build build-baseline
-cmake --build build-current
-
-# run both, back to back
-build-baseline/json_benchmarks --benchmark_repetitions=10 --benchmark_enable_random_interleaving=true \
-                               --benchmark_out=build-baseline/results.json --benchmark_out_format=json
-build-current/json_benchmarks --benchmark_repetitions=10 --benchmark_enable_random_interleaving=true \
-                              --benchmark_out=build-current/results.json --benchmark_out_format=json
-```
-
-Google Benchmark ships a tool to compare the two result files. It needs NumPy and SciPy:
-
-```sh
-python3 -m venv build-venv
-build-venv/bin/pip install numpy scipy
-build-venv/bin/python build-current/_deps/benchmark-src/tools/compare.py -a benchmarks build-baseline/results.json build-current/results.json
-```
-
-The tool's own `tools/requirements.txt` pins NumPy and SciPy versions that need Python 3.11 or later; with an older
-Python, unpinned versions work as well. In its output:
-
-- the `Time` and `CPU` columns are relative changes: `-0.35` means 35% faster, `+0.10` means 10% slower;
-- `_pvalue` lines report a Mann-Whitney U test of whether the two versions differ. It needs at least 9
-  repetitions, and a p-value below 0.05 means the difference is unlikely to be noise;
-- `OVERALL_GEOMEAN` summarizes all benchmarks;
-- `-a` shows only the aggregates, not every repetition.
-
-The header you compare with must support everything the benchmarks use. The current benchmarks build against 3.12.0.
-Only benchmarks present in both result files are compared, so for older releases, either filter the benchmarks or
-build that release's own `tests/benchmarks` against its own header.
-
-## Getting stable numbers
-
-- Build and run both versions on the same machine, one right after the other.
-- Keep the machine otherwise idle: no builds, no browser, and a laptop plugged in.
-- On Linux, set the CPU frequency governor to `performance`, e.g. `sudo cpupower frequency-set --governor performance`.
-  Google Benchmark prints a warning when frequency scaling is enabled. Pinning the process to a core
-  (`taskset -c 2 ...`) helps as well.
-- Use 10 or more repetitions with random interleaving, compare medians, and treat changes within the `_cv` as noise.
-
-## When to run them
-
-They are a manual step, not part of CI: shared CI runners vary more between runs than most of the effects measured.
-Run the comparison above before a release, comparing the previous release tag with `develop`, and for pull requests
-that claim to change performance.

tests/benchmarks/src/benchmarks.cpp

@@ -131,8 +131,7 @@ static void Dump(benchmark::State& state, const char* filename, int indent)
 
     while (state.KeepRunning())
     {
-        std::string output = j.dump(indent);
-        benchmark::DoNotOptimize(output);
+        j.dump(indent);
     }
 
     state.SetBytesProcessed(state.iterations() * j.dump(indent).size());

tests/fuzzing.md

@@ -79,3 +79,26 @@ the same `fuzzers` target as above and also relies on the `FUZZER_ENGINE` variab
 [build script](https://github.com/google/oss-fuzz/blob/master/projects/json/build.sh) for more information.
 
 In case the build at OSS-Fuzz fails, an issue will be created automatically.
+
+### Handling OSS-Fuzz reports
+
+OSS-Fuzz files the crashes it finds in its own [issue tracker](https://issues.oss-fuzz.com), not on GitHub. So that
+each report can be traced to the change that fixed it, and each fix to the report it answers, fixes follow these
+conventions:
+
+- **Reference the OSS-Fuzz issue in the pull request**, next to any GitHub issue it closes, as `OSS-Fuzz: <id>` (for
+  example, `OSS-Fuzz: 563659413`), and in the commit message. The ID alone does not disclose the crash. If the report
+  was triaged into a GitHub issue, link the OSS-Fuzz issue there too.
+- **Turn the reproducer into a unit test.** Download the testcase from the OSS-Fuzz report, reduce it if possible, and
+  add it as a regression test to the unit test of the affected format (e.g., `tests/src/unit-bjdata.cpp`), with a
+  comment naming the OSS-Fuzz issue. This way the input is checked by every CI run rather than only by OSS-Fuzz, and
+  it stays covered even if OSS-Fuzz later closes the report as not reproducible.
+- **Keep the fuzzer drivers and the unit tests in sync.** The round-trip checks of the UBJSON and BJData drivers are
+  also run on a fixed corpus in the unit tests (see `tests/src/round_trip_corpus.hpp` and the "round-trip invariants"
+  test cases), so a regression shows up in CI first. When a driver's checks change, change the unit tests with them.
+- **Record in the report whether the bug shipped.** OSS-Fuzz asks whether a crash was a short-lived regression or
+  affects a released version; answer it when the fix is merged, as it decides whether the fix needs a release note or
+  a security advisory (see the [security policy](../.github/SECURITY.md)).
+
+After the fix is merged, OSS-Fuzz re-runs the reproducer on its next build and marks the report as verified and
+closed. If it does not, the fix is incomplete.

tests/src/fuzzer-parse_bjdata.cpp

@@ -42,14 +42,23 @@ dump() serializes any non-finite double the same deterministic way (as JSON
 `null`, since JSON itself cannot represent NaN/Infinity), so comparing
 dumps is stable under exactly the same values that break operator==.
 
+The unit tests run the same checks on a fixed corpus (see the "BJData round-trip
+invariants" test case), so keep both in sync.
+
 The provided function `LLVMFuzzerTestOneInput` can be used in different fuzzer
 drivers.
 */
 
+#include <cassert>
 #include <iostream>
 #include <sstream>
 #include <nlohmann/json.hpp>
 
+// the round-trip checks below are assertions; NDEBUG would compile them away
+#ifdef NDEBUG
+    #error "the fuzzer drivers must be built without NDEBUG"
+#endif
+
 using json = nlohmann::json;
 
 // value-stable comparison for the round-trip checks below; see the note

tests/src/fuzzer-parse_bson.cpp

@@ -19,10 +19,16 @@ The provided function `LLVMFuzzerTestOneInput` can be used in different fuzzer
 drivers.
 */
 
+#include <cassert>
 #include <iostream>
 #include <sstream>
 #include <nlohmann/json.hpp>
 
+// the round-trip checks below are assertions; NDEBUG would compile them away
+#ifdef NDEBUG
+    #error "the fuzzer drivers must be built without NDEBUG"
+#endif
+
 using json = nlohmann::json;
 
 // see http://llvm.org/docs/LibFuzzer.html

tests/src/fuzzer-parse_cbor.cpp

@@ -19,10 +19,16 @@ The provided function `LLVMFuzzerTestOneInput` can be used in different fuzzer
 drivers.
 */
 
+#include <cassert>
 #include <iostream>
 #include <sstream>
 #include <nlohmann/json.hpp>
 
+// the round-trip checks below are assertions; NDEBUG would compile them away
+#ifdef NDEBUG
+    #error "the fuzzer drivers must be built without NDEBUG"
+#endif
+
 using json = nlohmann::json;
 
 // see http://llvm.org/docs/LibFuzzer.html

tests/src/fuzzer-parse_json.cpp

@@ -20,10 +20,16 @@ The provided function `LLVMFuzzerTestOneInput` can be used in different fuzzer
 drivers.
 */
 
+#include <cassert>
 #include <iostream>
 #include <sstream>
 #include <nlohmann/json.hpp>
 
+// the round-trip checks below are assertions; NDEBUG would compile them away
+#ifdef NDEBUG
+    #error "the fuzzer drivers must be built without NDEBUG"
+#endif
+
 using json = nlohmann::json;
 
 // see http://llvm.org/docs/LibFuzzer.html

tests/src/fuzzer-parse_msgpack.cpp

@@ -19,10 +19,16 @@ The provided function `LLVMFuzzerTestOneInput` can be used in different fuzzer
 drivers.
 */
 
+#include <cassert>
 #include <iostream>
 #include <sstream>
 #include <nlohmann/json.hpp>
 
+// the round-trip checks below are assertions; NDEBUG would compile them away
+#ifdef NDEBUG
+    #error "the fuzzer drivers must be built without NDEBUG"
+#endif
+
 using json = nlohmann::json;
 
 // see http://llvm.org/docs/LibFuzzer.html

tests/src/fuzzer-parse_ubjson.cpp

@@ -21,14 +21,23 @@ array data, it performs the following steps:
 - j4 = from_ubjson(vec3)
 - assert(j1 == j4)
 
+The unit tests run the same checks on a fixed corpus (see the "UBJSON round-trip
+invariants" test case), so keep both in sync.
+
 The provided function `LLVMFuzzerTestOneInput` can be used in different fuzzer
 drivers.
 */
 
+#include <cassert>
 #include <iostream>
 #include <sstream>
 #include <nlohmann/json.hpp>
 
+// the round-trip checks below are assertions; NDEBUG would compile them away
+#ifdef NDEBUG
+    #error "the fuzzer drivers must be built without NDEBUG"
+#endif
+
 using json = nlohmann::json;
 
 // see http://llvm.org/docs/LibFuzzer.html

tests/src/round_trip_corpus.hpp

@@ -0,0 +1,213 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  JSON for Modern C++ (supporting code)
+// |  |  |__   |  |  | | | |  version 3.12.0
+// |_____|_____|_____|_|___|  https://github.com/nlohmann/json
+//
+// SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>
+// SPDX-License-Identifier: MIT
+
+#pragma once
+
+#include <cmath> // nan
+#include <cstddef> // size_t
+#include <cstdint> // int32_t, int64_t, uint32_t, uint64_t
+#include <limits> // numeric_limits
+#include <random> // mt19937
+#include <string> // string, to_string
+#include <utility> // move
+#include <vector> // vector
+
+#include <nlohmann/json.hpp>
+
+// Values for the round-trip property tests of the UBJSON and BJData writers.
+//
+// The fuzzer drivers (tests/src/fuzzer-parse_ubjson.cpp and
+// fuzzer-parse_bjdata.cpp) check that anything the library parses can be
+// serialized, parsed back, and serialized again without loss. Those checks
+// only run at OSS-Fuzz, so a regression used to surface days later as an
+// external report. The unit tests run the same checks on this corpus in CI.
+//
+// The corpus is deterministic: std::mt19937's output sequence is fixed by
+// the standard, and it is used directly rather than through a distribution
+// (whose results are implementation-defined).
+namespace utils
+{
+
+class round_trip_corpus
+{
+  public:
+    using json = nlohmann::json;
+
+    static std::vector<json> values()
+    {
+        round_trip_corpus corpus;
+        return corpus.build();
+    }
+
+    // whether a value contains a binary value, which a BJData or UBJSON round
+    // trip may turn into an array of integers
+    static bool contains_binary(const json& j)
+    {
+        if (j.is_binary())
+        {
+            return true;
+        }
+        if (j.is_structured())
+        {
+            for (const auto& element : j)
+            {
+                if (contains_binary(element))
+                {
+                    return true;
+                }
+            }
+        }
+        return false;
+    }
+
+  private:
+    std::vector<json> atoms;
+    // a fixed seed is the point: the corpus must be the same in every run
+    std::mt19937 generator{42}; // NOLINT(cert-msc32-c,cert-msc51-cpp,bugprone-random-generator-seed)
+
+    round_trip_corpus()
+        : atoms
+    {
+        nullptr, true, false,
+        // integers at the boundaries of every UBJSON/BJData integer type
+        0, 1, -1, 127, 128, 255, 256, -128, -129,
+        32767, 32768, 65535, 65536, -32768, -32769,
+        (std::numeric_limits<std::int32_t>::min)(), (std::numeric_limits<std::int32_t>::max)(),
+        (std::numeric_limits<std::uint32_t>::max)(),
+        (std::numeric_limits<std::int64_t>::min)(), (std::numeric_limits<std::int64_t>::max)(),
+        static_cast<std::uint64_t>((std::numeric_limits<std::int64_t>::max)()) + 1u,
+        (std::numeric_limits<std::uint64_t>::max)(),
+        // floating-point numbers, including non-finite ones
+        0.0, -0.0, 1.5, -2.25, 3.4e38, (std::numeric_limits<double>::max)(),
+        std::nan(""), std::numeric_limits<double>::infinity(), -std::numeric_limits<double>::infinity(),
+        // strings, including a non-ASCII one and one longer than 255 bytes
+        "", "a", "\xC3\xA4", std::string(300, 'x'),
+        // binary values with and without subtype
+        json::binary({}), json::binary({1, 2, 255}), json::binary({0x80, 0x7F}, 42), json::binary({1}, 0)
+    }
+    {}
+
+    std::vector<json> build()
+    {
+        std::vector<json> result = atoms;
+
+        // each atom inside containers, including homogeneous ones that the
+        // writers encode as optimized (typed) containers
+        result.emplace_back(json::array());
+        result.emplace_back(json::object());
+        for (const auto& atom : atoms)
+        {
+            result.push_back(json::array({atom}));
+            result.push_back(json::array({atom, atom, atom}));
+            result.push_back(json::array({json::array({atom})}));
+            result.push_back(json::object({{"key", atom}}));
+        }
+        result.push_back(json::array({1, 1.5}));
+        result.push_back(json::array({-1, 255}));
+        result.push_back(json::array({"a", "b"}));
+
+        // deep, but well below any recursion or depth limit
+        json nested_array = 1;
+        json nested_object = 1;
+        for (int i = 0; i < 300; ++i)
+        {
+            nested_array = json::array({nested_array});
+            nested_object = json::object({{"key", nested_object}});
+        }
+        result.push_back(nested_array);
+        result.push_back(nested_object);
+
+        add_annotated_arrays(result);
+        add_random_values(result);
+        return result;
+    }
+
+    // objects in the JData annotated array format, which the BJData writer
+    // encodes as ND-arrays when the annotation describes a packed array, and
+    // as plain objects otherwise (see #5398, #5399, #5403, #5404, and #5542)
+    static void add_annotated_arrays(std::vector<json>& result)
+    {
+        const std::vector<json> types =
+        {
+            "uint8", "int8", "uint16", "int16", "uint32", "int32", "uint64", "int64",
+            "single", "double", "char", "byte", "bool", "unknown", 5, nullptr
+        };
+        const std::vector<json> sizes =
+        {
+            json::array(), {3}, {1, 3}, {3, 1}, {2, 3}, {2, 0}, {0, 2}, {2, 2, 2}, {-1, 2}, {2, 1.5},
+            "3", 3, nullptr, json::binary({})
+        };
+        const std::vector<json> data =
+        {
+            nullptr, 5, "s", json::object({{"a", 1}}), json::array(),
+            {1, 2, 3}, {1, 2, 3, 4, 5, 6}, {1, 2, 3, 4, 5, 6, 7, 8},
+            {1.5, 2.5, 3.5, 4.5, 5.5, 6.5}, {300, -300, 70000, -70000, 1, 2},
+            {"a", "b", "c", "d", "e", "f"}, {json::array({1, 2, 3}), json::array({4, 5, 6})}
+        };
+
+        for (const auto& type : types)
+        {
+            for (const auto& size : sizes)
+            {
+                for (const auto& d : data)
+                {
+                    result.push_back({{"_ArrayType_", type}, {"_ArraySize_", size}, {"_ArrayData_", d}});
+                }
+            }
+        }
+
+        // incomplete annotations and annotations with an extra key
+        result.push_back({{"_ArraySize_", {2, 3}}, {"_ArrayData_", {1, 2, 3, 4, 5, 6}}});
+        result.push_back({{"_ArrayType_", "uint8"}, {"_ArrayData_", {1, 2, 3, 4, 5, 6}}});
+        result.push_back({{"_ArrayType_", "uint8"}, {"_ArraySize_", {2, 3}}});
+        result.push_back({{"_ArrayType_", "uint8"}, {"_ArraySize_", {2, 3}}, {"_ArrayData_", {1, 2, 3, 4, 5, 6}}, {"extra", 1}});
+    }
+
+    // random containers of atoms, both homogeneous and mixed
+    void add_random_values(std::vector<json>& result)
+    {
+        for (int i = 0; i < 1000; ++i)
+        {
+            result.push_back(random_value(0));
+        }
+    }
+
+    std::size_t random_below(std::size_t bound)
+    {
+        return generator() % bound;
+    }
+
+    json random_value(int depth)
+    {
+        const auto kind = random_below(10);
+        if (depth > 3 || kind < 5)
+        {
+            return atoms[random_below(atoms.size())];
+        }
+
+        json result = kind < 8 ? json::array() : json::object();
+        const auto count = random_below(5);
+        const bool homogeneous = random_below(2) == 0;
+        const json fixed = atoms[random_below(atoms.size())];
+        for (std::size_t i = 0; i < count; ++i)
+        {
+            json element = homogeneous ? fixed : random_value(depth + 1);
+            if (result.is_array())
+            {
+                result.push_back(std::move(element));
+            }
+            else
+            {
+                result[std::to_string(i)] = std::move(element);
+            }
+        }
+        return result;
+    }
+};
+
+} // namespace utils

tests/src/unit-allocator.cpp

@@ -216,6 +216,57 @@ TEST_CASE("controlled bad_alloc")
             CHECK_THROWS_AS(my_json(s), std::bad_alloc&);
             next_construct_fails = false;
         }
+
+        SECTION("basic_json(const basic_json&) of a deeply nested value (#5387)")
+        {
+            // Copying a value nested deeper than the descent bound builds the
+            // copy from the top down: every value whose own copy has not been
+            // made yet stays a null value until it is. Failing an allocation
+            // part-way through is what proves such a half-built copy can still
+            // be destroyed.
+            //
+            // Which path the failure lands in depends on the build: the first
+            // allocation of a copy belongs to the outermost level, so here it
+            // is the descending one. Built with JSON_NO_THREAD_LOCAL - as the
+            // ci_test_no_thread_local target builds the whole suite - no
+            // descent is made at all and the very same failure lands in the
+            // iterative path instead, part-way through its worklist.
+            const auto check_deep_copy = [](bool objects)
+            {
+                CAPTURE(objects);
+
+                next_construct_fails = false;
+
+                // deeper than the 128 levels the copy constructor descends into
+                const std::size_t depth = 300;
+
+                my_json j = 1;
+                for (std::size_t i = 0; i < depth; ++i)
+                {
+                    if (objects)
+                    {
+                        my_json wrapper = my_json::object();
+                        wrapper["a"] = std::move(j);
+                        j = std::move(wrapper);
+                    }
+                    else
+                    {
+                        j = my_json::array({std::move(j)});
+                    }
+                }
+
+                // NOLINTNEXTLINE(performance-unnecessary-copy-initialization): the copy is what is tested
+                CHECK_NOTHROW(my_json(j));
+
+                next_construct_fails = true;
+                // NOLINTNEXTLINE(performance-unnecessary-copy-initialization): the copy is what is tested
+                CHECK_THROWS_AS(my_json(j), std::bad_alloc&);
+                next_construct_fails = false;
+            };
+
+            check_deep_copy(false);
+            check_deep_copy(true);
+        }
     }
 }
 

tests/src/unit-bjdata.cpp

@@ -19,6 +19,7 @@ using nlohmann::json;
 #include <fstream>
 #include <set>
 #include "make_test_data_available.hpp"
+#include "round_trip_corpus.hpp"
 #include "test_utils.hpp"
 
 namespace
@@ -2604,25 +2605,25 @@ TEST_CASE("BJData")
                 // that still round-trips.
 
                 // string data declared as a uint64 array
-                json const j_str = json({{"_ArrayType_", "uint64"}, {"_ArraySize_", {1}}, {"_ArrayData_", {"pointer"}}});
+                json const j_str = json({{"_ArrayType_", "uint64"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {"pointer", "value"}}});
                 const auto out_str = json::to_bjdata(j_str);
                 CHECK(out_str.at(0) == '{');
                 CHECK(json::from_bjdata(out_str) == j_str);
 
                 // integer data declared as a double array
-                json const j_float = json({{"_ArrayType_", "double"}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, 2}}});
+                json const j_float = json({{"_ArrayType_", "double"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {1, 2}}});
                 const auto out_float = json::to_bjdata(j_float);
                 CHECK(out_float.at(0) == '{');
                 CHECK(json::from_bjdata(out_float) == j_float);
 
                 // a non-integer shape entry is likewise not treated as an ndarray
-                json const j_size = json({{"_ArrayType_", "uint8"}, {"_ArraySize_", {"x"}}, {"_ArrayData_", {1}}});
+                json const j_size = json({{"_ArrayType_", "uint8"}, {"_ArraySize_", {"x", 1}}, {"_ArrayData_", {1}}});
                 const auto out_size = json::to_bjdata(j_size);
                 CHECK(out_size.at(0) == '{');
                 CHECK(json::from_bjdata(out_size) == j_size);
 
                 // a negative shape entry is not a usable dimension either
-                json const j_neg = json::parse(R"({"_ArrayType_":"uint8","_ArraySize_":[-1],"_ArrayData_":[1]})");
+                json const j_neg = json::parse(R"({"_ArrayType_":"uint8","_ArraySize_":[-1,1],"_ArrayData_":[1]})");
                 const auto out_neg = json::to_bjdata(j_neg);
                 CHECK(out_neg.at(0) == '{');
                 CHECK(json::from_bjdata(out_neg) == j_neg);
@@ -2657,14 +2658,14 @@ TEST_CASE("BJData")
                 }
 
                 // negative values under a signed type behave the same way
-                const auto from_neg = json::to_bjdata(json::parse(R"({"_ArrayType_":"int32","_ArraySize_":[2],"_ArrayData_":[-5,7]})"));
+                const auto from_neg = json::to_bjdata(json::parse(R"({"_ArrayType_":"int32","_ArraySize_":[2,1],"_ArrayData_":[-5,7]})"));
                 CHECK(from_neg.at(0) == '[');
-                CHECK(from_neg == json::to_bjdata(json({{"_ArrayType_", "int32"}, {"_ArraySize_", {2}}, {"_ArrayData_", {-5, 7}}})));
+                CHECK(from_neg == json::to_bjdata(json({{"_ArrayType_", "int32"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {-5, 7}}})));
 
                 // and so do the floating point types
-                const auto from_float = json::to_bjdata(json::parse(R"({"_ArrayType_":"double","_ArraySize_":[2],"_ArrayData_":[1.5,2.5]})"));
+                const auto from_float = json::to_bjdata(json::parse(R"({"_ArrayType_":"double","_ArraySize_":[2,1],"_ArrayData_":[1.5,2.5]})"));
                 CHECK(from_float.at(0) == '[');
-                CHECK(from_float == json::to_bjdata(json({{"_ArrayType_", "double"}, {"_ArraySize_", {2}}, {"_ArrayData_", {1.5, 2.5}}})));
+                CHECK(from_float == json::to_bjdata(json({{"_ArrayType_", "double"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {1.5, 2.5}}})));
             }
 
             SECTION("optimized ndarray (type and vector-size as 1D array)")
@@ -2835,7 +2836,7 @@ TEST_CASE("BJData")
                 // a single dimension that does not fit into std::size_t is
                 // rejected for the same reason (only observable where
                 // std::size_t is narrower than 64 bit)
-                json j_huge = json({{"_ArrayData_", json::array()}, {"_ArraySize_", {18446744073709551615ull}}, {"_ArrayType_", "uint8"}});
+                json j_huge = json({{"_ArrayData_", json::array()}, {"_ArraySize_", {18446744073709551615ull, 2}}, {"_ArrayType_", "uint8"}});
                 CHECK(json::from_bjdata(json::to_bjdata(j_huge), true, true) == j_huge);
 
                 // a well-formed ndarray is still encoded as one
@@ -2867,6 +2868,21 @@ TEST_CASE("BJData")
                 const auto out_num = json::to_bjdata(j_num);
                 CHECK(out_num.at(0) == '{');
                 CHECK(json::from_bjdata(out_num) == j_num);
+
+                // OSS-Fuzz issue 474400817: an empty object _ArraySize_ was
+                // written as the ND-array header length, which from_bjdata()
+                // could not read back
+                const std::vector<uint8_t> input =
+                {
+                    '[', '{', 'U', 11, '_', 'A', 'r', 'r', 'a', 'y', 'D', 'a', 't', 'a', '_', 'Z',
+                    'U', 11, '_', 'A', 'r', 'r', 'a', 'y', 'T', 'y', 'p', 'e', '_', 'S', 'i', 5, 'i', 'n', 't', '1', '6',
+                    'U', 11, '_', 'A', 'r', 'r', 'a', 'y', 'S', 'i', 'z', 'e', '_', '{', '}', '}', ']'
+                };
+                const json j1 = json::from_bjdata(input);
+                CHECK(j1 == json::parse(R"([{"_ArrayType_":"int16","_ArraySize_":{},"_ArrayData_":null}])"));
+                json j2;
+                CHECK_NOTHROW(j2 = json::from_bjdata(json::to_bjdata(j1, false, false)));
+                CHECK(j2 == j1);
             }
 
             SECTION("ndarray with out-of-range _ArrayData_ elements stays as object")
@@ -2878,42 +2894,116 @@ TEST_CASE("BJData")
                 // object encoding that still round-trips (see GitHub issue #5403)
 
                 // an unsigned element that does not fit uint8
-                json const j_uint8 = json({{"_ArrayType_", "uint8"}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, 256}}});
+                json const j_uint8 = json({{"_ArrayType_", "uint8"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {1, 256}}});
                 const auto out_uint8 = json::to_bjdata(j_uint8);
                 CHECK(out_uint8.at(0) == '{');
                 CHECK(json::from_bjdata(out_uint8) == j_uint8);
 
                 // a signed element that does not fit int8
-                json const j_int8 = json({{"_ArrayType_", "int8"}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, 200}}});
+                json const j_int8 = json({{"_ArrayType_", "int8"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {1, 200}}});
                 const auto out_int8 = json::to_bjdata(j_int8);
                 CHECK(out_int8.at(0) == '{');
                 CHECK(json::from_bjdata(out_int8) == j_int8);
 
                 // a negative element is likewise out of range for an
                 // unsigned _ArrayType_
-                json const j_uint16_neg = json({{"_ArrayType_", "uint16"}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, -1}}});
+                json const j_uint16_neg = json({{"_ArrayType_", "uint16"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {1, -1}}});
                 const auto out_uint16_neg = json::to_bjdata(j_uint16_neg);
                 CHECK(out_uint16_neg.at(0) == '{');
                 CHECK(json::from_bjdata(out_uint16_neg) == j_uint16_neg);
 
                 // a double element that overflows to infinity when narrowed
                 // to the "single" (float) precision named by _ArrayType_
-                json const j_single = json({{"_ArrayType_", "single"}, {"_ArraySize_", {2}}, {"_ArrayData_", {1.5, 1e40}}});
+                json const j_single = json({{"_ArrayType_", "single"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {1.5, 1e40}}});
                 const auto out_single = json::to_bjdata(j_single);
                 CHECK(out_single.at(0) == '{');
                 CHECK(json::from_bjdata(out_single) == j_single);
 
                 // in-range boundary values still use the compact ndarray encoding
-                json const j_uint8_ok = json({{"_ArrayType_", "uint8"}, {"_ArraySize_", {2}}, {"_ArrayData_", {0, 255}}});
-                CHECK(json::to_bjdata(j_uint8_ok) == std::vector<uint8_t>({'[', '$', 'U', '#', '[', 'i', 2, ']', 0, 255}));
+                json const j_uint8_ok = json({{"_ArrayType_", "uint8"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {0, 255}}});
+                CHECK(json::to_bjdata(j_uint8_ok) == std::vector<uint8_t>({'[', '$', 'U', '#', '[', 'i', 2, 'i', 1, ']', 0, 255}));
 
-                json const j_int8_ok = json({{"_ArrayType_", "int8"}, {"_ArraySize_", {2}}, {"_ArrayData_", {-128, 127}}});
-                CHECK(json::to_bjdata(j_int8_ok) == std::vector<uint8_t>({'[', '$', 'i', '#', '[', 'i', 2, ']', 0x80, 0x7F}));
+                json const j_int8_ok = json({{"_ArrayType_", "int8"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {-128, 127}}});
+                CHECK(json::to_bjdata(j_int8_ok) == std::vector<uint8_t>({'[', '$', 'i', '#', '[', 'i', 2, 'i', 1, ']', 0x80, 0x7F}));
 
-                json const j_single_ok = json({{"_ArrayType_", "single"}, {"_ArraySize_", {1}}, {"_ArrayData_", {1.5}}});
+                json const j_single_ok = json({{"_ArrayType_", "single"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {1.5, -1.5}}});
                 const auto out_single_ok = json::to_bjdata(j_single_ok);
                 CHECK(out_single_ok.at(0) == '[');
-                CHECK(json::from_bjdata(out_single_ok) == json({1.5f}));
+                CHECK(json::from_bjdata(out_single_ok) == json({{"_ArrayType_", "single"}, {"_ArraySize_", {2, 1}}, {"_ArrayData_", {1.5f, -1.5f}}}));
+            }
+
+            SECTION("ndarray that would not be read back as an annotated object stays as object")
+            {
+                // the reader only restores an annotated object from an ND-array
+                // with at least two non-zero dimensions that is not a 1xN row
+                // vector; any other shape is read back as a plain array. Writing
+                // such an object as an ND-array would drop its annotation, so it
+                // falls back to a plain object encoding that round-trips.
+                for (const char* text :
+                        {
+                            R"({"_ArrayType_":"int16","_ArraySize_":[],"_ArrayData_":[]})",
+                            R"({"_ArrayType_":"int16","_ArraySize_":[2],"_ArrayData_":[1,2]})",
+                            R"({"_ArrayType_":"int16","_ArraySize_":[1,2],"_ArrayData_":[1,2]})",
+                            R"({"_ArrayType_":"int16","_ArraySize_":[0],"_ArrayData_":[]})",
+                            R"({"_ArrayType_":"int16","_ArraySize_":[2,0],"_ArrayData_":[]})",
+                            R"({"_ArrayType_":"int16","_ArraySize_":[0,2],"_ArrayData_":[]})"
+                        })
+                {
+                    CAPTURE(text);
+                    const json j = json::parse(text);
+                    for (const bool use_size :
+                            {
+                                false, true
+                            })
+                    {
+                        const auto out = json::to_bjdata(j, use_size, use_size);
+                        CHECK(out.at(0) == '{');
+                        CHECK(json::from_bjdata(out) == j);
+                    }
+                }
+
+                // a genuine ND-array still uses the compact encoding and round-trips
+                const json j_2d = json::parse(R"({"_ArrayType_":"int16","_ArraySize_":[2,1],"_ArrayData_":[1,2]})");
+                const auto out_2d = json::to_bjdata(j_2d);
+                CHECK(out_2d.at(0) == '[');
+                CHECK(json::from_bjdata(out_2d) == j_2d);
+            }
+
+            SECTION("ndarray with non-array _ArrayData_ stays as object")
+            {
+                // the elements are written from _ArrayData_ as a flat list, so it
+                // has to be an array: null has size 0, any other scalar has size 1,
+                // and iterating an object visits its values, so each of these could
+                // match the dimensions and be encoded as an unrelated ND-array
+                for (const char* text :
+                        {
+                            R"({"_ArrayType_":"int16","_ArraySize_":[2,1],"_ArrayData_":null})",
+                            R"({"_ArrayType_":"int16","_ArraySize_":[2,1],"_ArrayData_":{"a":1,"b":2}})",
+                            R"({"_ArrayType_":"int16","_ArraySize_":[1],"_ArrayData_":5})",
+                            R"({"_ArrayType_":"int16","_ArraySize_":[],"_ArrayData_":null})"
+                        })
+                {
+                    CAPTURE(text);
+                    const json j = json::parse(text);
+                    const auto out = json::to_bjdata(j);
+                    CHECK(out.at(0) == '{');
+                    CHECK(json::from_bjdata(out) == j);
+                }
+
+                // OSS-Fuzz issue 563659413: an empty binary _ArraySize_ is written
+                // as a plain object and read back as an empty array, after which
+                // the object with a null _ArrayData_ was encoded as an empty
+                // ND-array and re-read as [], so a second round trip lost the value
+                const std::vector<uint8_t> input =
+                {
+                    '{', 'U', 11, '_', 'A', 'r', 'r', 'a', 'y', 'D', 'a', 't', 'a', '_', 'Z',
+                    'U', 11, '_', 'A', 'r', 'r', 'a', 'y', 'T', 'y', 'p', 'e', '_', 'S', 'i', 5, 'i', 'n', 't', '1', '6',
+                    'U', 11, '_', 'A', 'r', 'r', 'a', 'y', 'S', 'i', 'z', 'e', '_', '[', '$', 'B', '#', '[', ']', '}'
+                };
+                const json j1 = json::from_bjdata(input);
+                const json j2 = json::from_bjdata(json::to_bjdata(j1, false, false));
+                CHECK(j2 == json::parse(R"({"_ArrayType_":"int16","_ArraySize_":[],"_ArrayData_":null})"));
+                CHECK(json::from_bjdata(json::to_bjdata(j2, false, false)) == j2);
             }
 
             SECTION("ndarray with _ArrayType_ \"byte\" is gated by the BJData draft version")
@@ -4199,6 +4289,93 @@ TEST_CASE("BJData use_type requires use_size")
     }
 }
 
+TEST_CASE("BJData round-trip invariants")
+{
+    // This checks what the parse_bjdata_fuzzer driver checks (see
+    // tests/src/fuzzer-parse_bjdata.cpp), so that a regression shows up in CI
+    // rather than as an OSS-Fuzz report: every value from_bjdata() returns
+    // (j1) can be serialized with any combination of options, the result can
+    // be parsed back (j2), and serializing j2 again with the same options
+    // yields a value-equal result.
+    //
+    // Beyond the driver, this also checks that j2 equals j1 and that
+    // serializing j2 reproduces the exact bytes, both except for values that
+    // contain a binary value: a binary value is only written as a binary
+    // value with Draft 3's optimized binary array, and otherwise read back as
+    // an array of integers, for which the writer may choose different (but
+    // equally valid) type markers when it is serialized again (see #5494).
+    //
+    // Values are compared with dump() rather than operator==, because a NaN
+    // never compares equal to itself.
+    struct options
+    {
+        bool use_size;
+        bool use_type;
+        json::bjdata_version_t version;
+    };
+    const std::vector<options> all_options =
+    {
+        {false, false, json::bjdata_version_t::draft2},
+        {true, false, json::bjdata_version_t::draft2},
+        {true, true, json::bjdata_version_t::draft2},
+        {false, false, json::bjdata_version_t::draft3},
+        {true, false, json::bjdata_version_t::draft3},
+        {true, true, json::bjdata_version_t::draft3},
+    };
+
+    for (const auto& j0 : utils::round_trip_corpus::values())
+    {
+        // turn the corpus value into a value as from_bjdata() returns it
+        for (const auto& initial : all_options)
+        {
+            const json j1 = json::from_bjdata(json::to_bjdata(j0, initial.use_size, initial.use_type, initial.version));
+            const bool has_binary = utils::round_trip_corpus::contains_binary(j1);
+
+            for (const auto& o : all_options)
+            {
+                INFO("j1 = " << j1.dump() << ", use_size = " << o.use_size << ", use_type = " << o.use_type
+                     << ", draft3 = " << (o.version == json::bjdata_version_t::draft3));
+
+                const std::vector<std::uint8_t> vec = json::to_bjdata(j1, o.use_size, o.use_type, o.version);
+                json j2;
+                // anything the library writes must be parsable by the library
+                REQUIRE_NOTHROW(j2 = json::from_bjdata(vec));
+                const std::vector<std::uint8_t> vec2 = json::to_bjdata(j2, o.use_size, o.use_type, o.version);
+                CHECK(json::from_bjdata(vec2).dump() == j2.dump());
+
+                if (!has_binary)
+                {
+                    CHECK(j2.dump() == j1.dump());
+                    CHECK(vec2 == vec);
+                }
+            }
+        }
+    }
+}
+
+TEST_CASE("BJData round trip of a binary value is value-stable, not byte-stable")
+{
+    // OSS-Fuzz issue 474480402: a Draft 3 optimized binary array is read as a
+    // binary value, which to_bjdata() writes in the default Draft 2 mode as a
+    // plain array of uint8 numbers. That is read back as an array of numbers,
+    // for which the writer then picks the smallest type marker, int8 ('i'),
+    // so re-serializing changes the bytes, but not the value. This is the
+    // exception described in the "Round trips" note of the BJData
+    // documentation, and why the fuzzer checks value stability (see #5494).
+    const std::vector<uint8_t> input = {'[', '$', 'B', '#', 'U', 1, 0x20};
+    const json j1 = json::from_bjdata(input);
+    CHECK(j1 == json::binary({0x20}));
+
+    const std::vector<uint8_t> vec = json::to_bjdata(j1, false, false);
+    CHECK(vec == std::vector<uint8_t>({'[', 'U', 0x20, ']'}));
+    const json j2 = json::from_bjdata(vec);
+    CHECK(j2 == json::array({0x20}));
+
+    const std::vector<uint8_t> vec2 = json::to_bjdata(j2, false, false);
+    CHECK(vec2 == std::vector<uint8_t>({'[', 'i', 0x20, ']'}));
+    CHECK(json::from_bjdata(vec2) == j2);
+}
+
 TEST_CASE("BJData roundtrips" * doctest::skip())
 {
     SECTION("input from self-generated BJData files")

tests/src/unit-brace-init-copy-semantics.cpp

@@ -0,0 +1,167 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  JSON for Modern C++ (supporting code)
+// |  |  |__   |  |  | | | |  version 3.12.0
+// |_____|_____|_____|_|___|  https://github.com/nlohmann/json
+//
+// SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>
+// SPDX-License-Identifier: MIT
+
+#include "doctest_compatibility.h"
+
+// This file tests the opt-in JSON_BRACE_INIT_COPY_SEMANTICS, so it defines the
+// macro itself rather than relying on a -D flag, and runs in every build.
+#ifdef JSON_BRACE_INIT_COPY_SEMANTICS
+    #undef JSON_BRACE_INIT_COPY_SEMANTICS
+#endif
+
+#define JSON_BRACE_INIT_COPY_SEMANTICS 1
+
+#include <nlohmann/json.hpp>
+using nlohmann::json;
+
+#include <array>
+#include <list>
+#include <map>
+#include <string>
+#include <tuple>
+#include <utility>
+#include <vector>
+
+#define STRINGIZE_EX(x) #x
+#define STRINGIZE(x) STRINGIZE_EX(x)
+
+TEST_CASE("JSON_BRACE_INIT_COPY_SEMANTICS")
+{
+    SECTION("the macro is part of the ABI tag")
+    {
+        const std::string ns = STRINGIZE(NLOHMANN_JSON_NAMESPACE);
+        // other tags may come before it, e.g. json_abi_ldvcmp_bics
+        CHECK(ns.find("_bics") != std::string::npos);
+    }
+
+    SECTION("single-element brace initialization copies the element (#5074)")
+    {
+        json const j_obj = {{"key", "value"}, {"num", 42}};
+        json const j_arr = {1, 2, 3};
+
+        // object: brace init copies instead of wrapping
+        json const j1{j_obj};
+        CHECK(j1.is_object());
+        CHECK(j1 == j_obj);
+
+        // array: brace init copies instead of wrapping
+        json const j2{j_arr};
+        CHECK(j2.is_array());
+        CHECK(j2.size() == 3);
+        CHECK(j2 == j_arr);
+
+        // this applies to any single element, not only to JSON values
+        json const j3{true};
+        CHECK(j3.is_boolean());
+
+        json const j4{42};
+        CHECK(j4.is_number_integer());
+
+        json const j5 = {1};
+        CHECK(j5 == 1);
+
+        json const j6 = {"text"};
+        CHECK(j6 == "text");
+
+        json const j7 = {{1, 2}};
+        CHECK(j7 == json::array({1, 2}));
+    }
+
+    SECTION("what the macro does not change")
+    {
+        // lists with more than one element are unaffected
+        json const j1 = {1, 2};
+        CHECK(j1.is_array());
+        CHECK(j1.size() == 2);
+
+        // a single [string, value] pair still describes an object
+        json const j2 = {{"key", "value"}};
+        CHECK(j2.is_object());
+        CHECK(j2["key"] == "value");
+
+        // json::array() always creates an array
+        json const j3 = json::array({1});
+        CHECK(j3.is_array());
+        CHECK(j3.size() == 1);
+        CHECK(j3[0] == 1);
+
+        json const j_obj = {{"key", "value"}};
+        json const j4 = json::array({j_obj});
+        CHECK(j4.is_array());
+        CHECK(j4.size() == 1);
+        CHECK(j4[0] == j_obj);
+    }
+
+    SECTION("conversions build the same values as without the macro")
+    {
+        SECTION("one-element std::tuple")
+        {
+            json const j1 = std::tuple<int> {5};
+            CHECK(j1.dump() == "[5]");
+            CHECK(std::get<0>(j1.get<std::tuple<int>>()) == 5);
+
+            json const j2 = std::tuple<std::string> {"text"};
+            CHECK(j2.dump() == "[\"text\"]");
+            CHECK(std::get<0>(j2.get<std::tuple<std::string>>()) == "text");
+
+            json const j3 = std::tuple<json> {json::array({1, 2})};
+            CHECK(j3.dump() == "[[1,2]]");
+
+            // as without the macro, a [string, value] pair becomes an object
+            // member (see the known limitation documented for std::pair)
+            json const j4 = std::tuple<std::pair<std::string, int>> {{"a", 1}};
+            CHECK(j4.dump() == "{\"a\":1}");
+        }
+
+        SECTION("tuples with more elements")
+        {
+            json const j1 = std::tuple<int, std::string> {1, "a"};
+            CHECK(j1.dump() == "[1,\"a\"]");
+
+            json const j2 = std::tuple<> {};
+            CHECK(j2.dump() == "[]");
+        }
+
+        SECTION("one-element containers")
+        {
+            json const j1 = std::vector<int> {1};
+            CHECK(j1.dump() == "[1]");
+            CHECK(j1.get<std::vector<int>>() == std::vector<int> {1});
+
+            std::array<int, 1> const arr = {{1}};
+            json const j2 = arr;
+            CHECK(j2.dump() == "[1]");
+
+            json const j3 = std::list<std::string> {"a"};
+            CHECK(j3.dump() == "[\"a\"]");
+
+            json const j4 = std::map<std::string, int> {{"a", 1}};
+            CHECK(j4.dump() == "{\"a\":1}");
+
+            json const j5 = std::map<int, int> {{1, 2}};
+            CHECK(j5.dump() == "[[1,2]]");
+        }
+
+        SECTION("std::pair")
+        {
+            json const j = std::pair<int, int> {1, 2};
+            CHECK(j.dump() == "[1,2]");
+            CHECK((j.get<std::pair<int, int>>() == std::pair<int, int> {1, 2}));
+        }
+
+        SECTION("items()")
+        {
+            json j_obj = {{"key", 1}};
+            for (const auto& el : j_obj.items())
+            {
+                json const j = el;
+                CHECK(j.dump() == "{\"key\":1}");
+            }
+        }
+    }
+}

tests/src/unit-deserialization.cpp

@@ -25,6 +25,7 @@ using nlohmann::json;
 #include <iostream>
 #include <iterator>
 #include <sstream>
+#include <string>
 #include <valarray>
 
 #if defined(_WIN32)
@@ -1233,6 +1234,59 @@ TEST_CASE("deserialization")
         }
     }
 
+    SECTION("stream position after extraction without JSON_PRECISE_STREAM_POSITION (#5340)")
+    {
+        // By default, the character that terminates a number is consumed, so
+        // the stream is left one byte too far after a number (and only after a
+        // number). JSON_PRECISE_STREAM_POSITION changes this; see
+        // unit-precise-stream-position.cpp. These checks pin the default.
+        const auto remaining = [](std::istream & is) -> std::string
+        {
+            return {std::istreambuf_iterator<char>(is), std::istreambuf_iterator<char>()};
+        };
+
+        SECTION("the character after a number is consumed")
+        {
+            std::istringstream ss("1true");
+            json j;
+            ss >> j;
+            CHECK(j == 1);
+            CHECK(remaining(ss) == "rue");
+        }
+
+        SECTION("the character after other values is not consumed")
+        {
+            std::istringstream ss("[1]true");
+            json j;
+            ss >> j;
+            CHECK(j == json::parse("[1]"));
+            CHECK(remaining(ss) == "true");
+        }
+
+        SECTION("comma-separated numbers can be read one by one")
+        {
+            std::istringstream ss("1,2,3");
+            json j1;
+            json j2;
+            json j3;
+            ss >> j1 >> j2 >> j3;
+            CHECK(j1 == 1);
+            CHECK(j2 == 2);
+            CHECK(j3 == 3);
+        }
+
+        SECTION("std::getline after a number skips the line break")
+        {
+            std::istringstream ss("42\nfoo");
+            json j;
+            std::string line;
+            ss >> j;
+            std::getline(ss, line);
+            CHECK(j == 42);
+            CHECK(line == "foo");
+        }
+    }
+
     // build with C++20
     // JSON_HAS_CPP_20
 #if defined(__cpp_char8_t)

tests/src/unit-diagnostic-positions.cpp

@@ -75,6 +75,72 @@ TEST_CASE("Better diagnostics with positions")
         CHECK(j.end_pos() == root.size());
     }
 
+    SECTION("copying keeps the positions of nested values (#5387)")
+    {
+        // Values nested deeper than the copy constructor's descent bound are
+        // copied without the call stack, on a path that has to carry the
+        // positions over itself; shallower ones copy their containers, which
+        // bring the positions along. Both sides of the bound are checked here.
+        const auto check_copy = [](std::size_t depth, bool objects)
+        {
+            CAPTURE(depth)
+            CAPTURE(objects)
+
+            const std::string opening = objects ? R"({"a":)" : "[";
+            const std::string closing = objects ? "}" : "]";
+
+            std::string text;
+            for (std::size_t i = 0; i < depth; ++i)
+            {
+                text += opening;
+            }
+            text += "12";
+            for (std::size_t i = 0; i < depth; ++i)
+            {
+                text += closing;
+            }
+
+            const json original = json::parse(text);
+            const json copy(original); // NOLINT(performance-unnecessary-copy-initialization)
+
+            const json* o = &original;
+            const json* c = &copy;
+            for (std::size_t level = 0; level <= depth; ++level)
+            {
+                CAPTURE(level)
+                REQUIRE(c->start_pos() == o->start_pos());
+                REQUIRE(c->end_pos() == o->end_pos());
+
+                if (level < depth)
+                {
+                    o = objects ? &o->at("a") : &o->at(0);
+                    c = objects ? &c->at("a") : &c->at(0);
+                }
+            }
+        };
+
+        const auto check_arrays = [&check_copy](std::size_t depth)
+        {
+            check_copy(depth, false);
+        };
+        const auto check_objects = [&check_copy](std::size_t depth)
+        {
+            check_copy(depth, true);
+        };
+
+        check_arrays(1);
+        check_arrays(127);
+        check_arrays(128);
+        check_arrays(129);
+        check_arrays(300);
+
+        check_objects(1);
+        check_objects(127);
+        check_objects(128);
+        check_objects(129);
+        check_objects(300);
+    }
+
     SECTION("JSON patch add to primitive parent (#4292)")
     {
         // the JSON Patch "add" target /foo/bar/baz has a string parent

tests/src/unit-diagnostics.cpp

@@ -274,6 +274,63 @@ TEST_CASE("Regression tests for extended diagnostics")
         CHECK(j1["string"] == "t");
     }
 
+    SECTION("Regression test for issue #5387 - copying keeps the parents of nested values")
+    {
+        // A value nested deeper than the copy constructor's descent bound is
+        // copied without the call stack. Every container that path creates has
+        // to have the parents of its children set, or the JSON Pointer in the
+        // diagnostic is cut short.
+        const std::size_t depth = 300;
+
+        SECTION("objects")
+        {
+            json j = "not a number";
+            std::string pointer;
+            for (std::size_t i = 0; i < depth; ++i)
+            {
+                j = json{{"a", j}};
+                pointer += "/a";
+            }
+
+            json const copy(j); // NOLINT(performance-unnecessary-copy-initialization)
+
+            const json* inner = &copy;
+            for (std::size_t i = 0; i < depth; ++i)
+            {
+                inner = &inner->at("a");
+            }
+
+            std::string const expected = "[json.exception.type_error.302] (" + pointer + ") type must be number, but is string";
+            int i = 0;
+            CHECK_THROWS_WITH_AS(i = inner->get<int>(), expected.c_str(), json::type_error);
+            CHECK(i == 0);
+        }
+
+        SECTION("arrays")
+        {
+            json j = "not a number";
+            std::string pointer;
+            for (std::size_t i = 0; i < depth; ++i)
+            {
+                j = json::array({j});
+                pointer += "/0";
+            }
+
+            json const copy(j); // NOLINT(performance-unnecessary-copy-initialization)
+
+            const json* inner = &copy;
+            for (std::size_t i = 0; i < depth; ++i)
+            {
+                inner = &inner->at(0);
+            }
+
+            std::string const expected = "[json.exception.type_error.302] (" + pointer + ") type must be number, but is string";
+            int i = 0;
+            CHECK_THROWS_WITH_AS(i = inner->get<int>(), expected.c_str(), json::type_error);
+            CHECK(i == 0);
+        }
+    }
+
     SECTION("Regression test - swap(array_t&)/swap(object_t&) must update JSON_DIAGNOSTICS parent pointers")
     {
         // swap(array_t&)
@@ -304,5 +361,154 @@ TEST_CASE("Regression tests for extended diagnostics")
             CHECK(p == o);
         }
     }
+
+    SECTION("Regression test - erase() and update() must keep JSON_DIAGNOSTICS parent pointers of ordered_json members")
+    {
+        // ordered_json keeps its members in a vector: erasing a member
+        // re-constructs all members after it in place, and adding a key may
+        // reallocate the vector; both reset the parent pointers of the members
+        // that were moved
+        using nlohmann::ordered_json;
+
+        const auto check_parents = [](const ordered_json & j)
+        {
+            // const access, so operator[] cannot repair the parent pointers
+            CHECK_THROWS_WITH_AS(j["z"]["x"].at(0), "[json.exception.type_error.304] (/z/x) cannot use at() with number", ordered_json::type_error);
+
+            // must not trigger assert_invariant() in a debug/assert-enabled build
+            ordered_json const copy = j; // NOLINT(performance-unnecessary-copy-initialization)
+            CHECK(copy == j);
+        };
+
+        // erase(key)
+        {
+            ordered_json j = {{"a", 1}, {"z", {{"x", 1}}}};
+            CHECK(j.erase("a") == 1);
+            check_parents(j);
+        }
+
+        // erase(iterator)
+        {
+            ordered_json j = {{"a", 1}, {"z", {{"x", 1}}}};
+            j.erase(j.begin());
+            check_parents(j);
+        }
+
+        // erase(iterator, iterator)
+        {
+            ordered_json j = {{"a", 1}, {"b", 2}, {"z", {{"x", 1}}}};
+            j.erase(j.begin(), j.find("z"));
+            check_parents(j);
+        }
+
+        // patch() removes via erase(iterator)
+        {
+            ordered_json j = {{"a", 1}, {"z", {{"x", 1}}}};
+            j.patch_inplace(ordered_json::parse(R"([{"op": "remove", "path": "/a"}])"));
+            check_parents(j);
+        }
+
+        // update(j)
+        {
+            ordered_json j = {{"z", {{"x", 1}}}};
+            j.update({{"a", 1}, {"b", 2}});
+            check_parents(j);
+        }
+
+        // update(j, true), the outer and the nested vector both grow
+        {
+            ordered_json j = {{"z", {{"x", 1}}}};
+            j.update({{"z", {{"y", 2}}}, {"a", 1}}, true);
+            check_parents(j);
+        }
+
+        // update(j, true) around its descent bound, where the nested vectors
+        // grow while the objects are merged without recursing
+        for (const std::size_t depth :
+                {
+                    nlohmann::detail::recursion_depth_limit() - 1, nlohmann::detail::recursion_depth_limit(), nlohmann::detail::recursion_depth_limit() + 2
+                })
+        {
+            ordered_json j = {{"z", {{"x", 1}}}};
+            ordered_json patch = {{"a", 1}, {"b", 2}, {"c", {{"d", 3}}}};
+            for (std::size_t i = 0; i < depth; ++i)
+            {
+                j = ordered_json{{"k", 0}, {"n", std::move(j)}};
+                patch = ordered_json{{"n", std::move(patch)}, {"l", 1}, {"m", 2}};
+            }
+            j.update(patch, true);
+
+            // must not trigger assert_invariant() on any level in a
+            // debug/assert-enabled build
+            ordered_json const copy = j; // NOLINT(performance-unnecessary-copy-initialization)
+            CHECK(copy == j);
+        }
+
+        // merge_patch() inserts "c" and removes "d" at /a/c, then inserts "e"
+        // at /a, which copies /a/c
+        {
+            auto j = ordered_json::parse(R"({"a": {"c": {"d": {}}}})");
+            j.merge_patch(ordered_json::parse(R"({"a": {"c": {"c": "s", "d": null}, "e": "s"}})"));
+            CHECK(j.dump() == R"({"a":{"c":{"c":"s"},"e":"s"}})");
+
+            auto const& constJ = j;
+#if JSON_DIAGNOSTIC_POSITIONS
+            CHECK_THROWS_WITH_AS(constJ["a"]["c"]["c"].at(0), "[json.exception.type_error.304] (/a/c/c) (bytes 18-21) cannot use at() with string", ordered_json::type_error);
+#else
+            CHECK_THROWS_WITH_AS(constJ["a"]["c"]["c"].at(0), "[json.exception.type_error.304] (/a/c/c) cannot use at() with string", ordered_json::type_error);
+#endif
+            ordered_json const copy = j;
+            CHECK(copy == j);
+        }
+    }
 }
 
+TEST_CASE("Better diagnostics past the descent bound of update() and merge_patch()")
+{
+    // Both merge objects nested more than detail::recursion_depth_limit()
+    // (128) levels deep without recursing; the values they add or replace
+    // there must still know their parents.
+    // The values are built rather than parsed, so that the expected messages
+    // carry no byte positions under JSON_DIAGNOSTIC_POSITIONS.
+    const std::size_t depth = 200;
+    json target = {{"x", 1}};
+    json patch = {{"y", 2}};
+    std::string path;
+    for (std::size_t i = 0; i < depth; ++i)
+    {
+        target = json{{"a", std::move(target)}};
+        patch = json{{"a", std::move(patch)}};
+        path += "/a";
+    }
+    const std::string expected_x = "[json.exception.type_error.304] (" + path + "/x) cannot use at() with number";
+    const std::string expected_y = "[json.exception.type_error.304] (" + path + "/y) cannot use at() with number";
+
+    SECTION("update()")
+    {
+        json j = target;
+        j.update(patch, true);
+
+        // walk down through const references, which leave m_parent alone
+        const json* p = &j;
+        for (std::size_t i = 0; i < depth; ++i)
+        {
+            p = &p->at("a");
+        }
+        CHECK_THROWS_WITH_AS(p->at("x").at(0), expected_x.c_str(), json::type_error);
+        CHECK_THROWS_WITH_AS(p->at("y").at(0), expected_y.c_str(), json::type_error);
+    }
+
+    SECTION("merge_patch()")
+    {
+        json j = target;
+        j.merge_patch(patch);
+
+        const json* p = &j;
+        for (std::size_t i = 0; i < depth; ++i)
+        {
+            p = &p->at("a");
+        }
+        CHECK_THROWS_WITH_AS(p->at("x").at(0), expected_x.c_str(), json::type_error);
+        CHECK_THROWS_WITH_AS(p->at("y").at(0), expected_y.c_str(), json::type_error);
+    }
+}

tests/src/unit-hash.cpp

@@ -13,6 +13,78 @@ using json = nlohmann::json;
 using ordered_json = nlohmann::ordered_json;
 
 #include <set>
+#include <string>
+
+namespace
+{
+// how detail::hash defines the hash of an array or object: the seeds of the
+// elements, combined in order. Recursive, so only usable on values nested a
+// few hundred levels deep - which is exactly what is needed to check that the
+// iterative path taken below detail::recursion_depth_limit() computes the same.
+template<typename BasicJsonType>
+std::size_t reference_hash(const BasicJsonType& j)
+{
+    using nlohmann::detail::combine;
+    using string_t = typename BasicJsonType::string_t;
+
+    if (!j.is_structured())
+    {
+        return std::hash<BasicJsonType> {}(j);
+    }
+
+    auto seed = combine(static_cast<std::size_t>(j.type()), j.size());
+    for (const auto& element : j.items())
+    {
+        if (j.is_object())
+        {
+            seed = combine(seed, std::hash<string_t> {}(element.key()));
+        }
+        seed = combine(seed, reference_hash(element.value()));
+    }
+    return seed;
+}
+
+// a value nested `depth` levels deep, with siblings on every level
+template<typename BasicJsonType>
+BasicJsonType nested(const std::size_t depth, const bool objects)
+{
+    BasicJsonType value = "leaf";
+    for (std::size_t i = 0; i < depth; ++i)
+    {
+        if (objects)
+        {
+            value = BasicJsonType{{"before", i}, {"nested", std::move(value)}, {"after", {i, "x"}}};
+        }
+        else
+        {
+            value = BasicJsonType::array({i, std::move(value), BasicJsonType::object({{"k", i}})});
+        }
+    }
+    return value;
+}
+
+std::string nested_text(const std::size_t depth, const bool objects)
+{
+    std::string text;
+    if (objects)
+    {
+        text.reserve((6 * depth) + 1);
+        for (std::size_t i = 0; i < depth; ++i)
+        {
+            text += "{\"a\":";
+        }
+        text += "1";
+        text.append(depth, '}');
+    }
+    else
+    {
+        text.assign(depth, '[');
+        text += "1";
+        text.append(depth, ']');
+    }
+    return text;
+}
+} // namespace
 
 TEST_CASE("hash<nlohmann::json>")
 {
@@ -111,3 +183,44 @@ TEST_CASE("hash<nlohmann::ordered_json>")
 
     CHECK(hashes.size() == 21);
 }
+
+TEST_CASE("hash of deeply nested values")
+{
+    SECTION("hashing past the descent bound computes the same values")
+    {
+        // every depth on either side of where the iterative path takes over
+        for (std::size_t depth = 0; depth <= (2 * nlohmann::detail::recursion_depth_limit()) + 10; ++depth)
+        {
+            CAPTURE(depth);
+            const auto arrays = nested<json>(depth, false);
+            const auto objects = nested<json>(depth, true);
+            const auto ordered = nested<ordered_json>(depth, true);
+            CHECK(std::hash<json> {}(arrays) == reference_hash(arrays));
+            CHECK(std::hash<json> {}(objects) == reference_hash(objects));
+            CHECK(std::hash<ordered_json> {}(ordered) == reference_hash(ordered));
+        }
+    }
+
+    SECTION("values nested too deeply for the call stack (#5545)")
+    {
+        // recursing once per level used to exhaust the call stack here; the
+        // values are only parsed and hashed, never copied or compared, since
+        // those recurse as well
+        const std::size_t depth = 100000;
+        for (const bool objects :
+                {
+                    false, true
+                })
+        {
+            CAPTURE(objects);
+            const auto text = nested_text(depth, objects);
+            const auto a = json::parse(text);
+            const auto b = json::parse(text);
+            CHECK(std::hash<json> {}(a) == std::hash<json> {}(b));
+
+            const auto c = ordered_json::parse(text);
+            const auto d = ordered_json::parse(text);
+            CHECK(std::hash<ordered_json> {}(c) == std::hash<ordered_json> {}(d));
+        }
+    }
+}

tests/src/unit-large_json.cpp

@@ -12,6 +12,7 @@
 using nlohmann::json;
 
 #include <algorithm>
+#include <string>
 
 TEST_CASE("tests on very large JSONs")
 {
@@ -27,3 +28,201 @@ TEST_CASE("tests on very large JSONs")
     }
 }
 
+namespace
+{
+
+// Descend a chain of single-element containers and return the value at its end,
+// reporting the number of levels traversed in @a depth.
+//
+// The values in the test case below are nested far deeper than the call stack
+// can follow, so they must not be inspected with operator== or dump(): both are
+// still recursive and would overflow the stack themselves.
+const json* innermost_value(const json& j, std::size_t& depth)
+{
+    const json* current = &j;
+    depth = 0;
+
+    while ((current->is_array() || current->is_object()) && !current->empty())
+    {
+        current = current->is_array()
+                  ? &current->front()
+                  : &current->begin().value();
+        ++depth;
+    }
+
+    return current;
+}
+
+} // namespace
+
+TEST_CASE("tests on deeply nested JSONs")
+{
+    // deep enough to exhaust the call stack, but small enough to stay cheap:
+    // parsing is iterative, so building the values below costs little
+    const std::size_t depth = 100000;
+
+    SECTION("issue #5387 - stack overflow in the copy constructor")
+    {
+        SECTION("array")
+        {
+            const json j = json::parse(std::string(depth, '[') + '0' + std::string(depth, ']'));
+
+            const json copy(j); // NOLINT(performance-unnecessary-copy-initialization): the copy is what is tested
+
+            std::size_t copy_depth = 0;
+            CHECK(*innermost_value(copy, copy_depth) == 0);
+            CHECK(copy_depth == depth);
+        }
+
+        SECTION("object")
+        {
+            std::string s;
+            s.reserve((6 * depth) + 1);
+            for (std::size_t i = 0; i < depth; ++i)
+            {
+                s += "{\"a\":";
+            }
+            s += '1';
+            s.append(depth, '}');
+
+            const json j = json::parse(s);
+
+            const json copy(j); // NOLINT(performance-unnecessary-copy-initialization): the copy is what is tested
+
+            std::size_t copy_depth = 0;
+            CHECK(*innermost_value(copy, copy_depth) == 1);
+            CHECK(copy_depth == depth);
+        }
+
+        SECTION("copy assignment")
+        {
+            // operator=(basic_json) takes its argument by value, so the deep
+            // copy happens in the copy constructor
+            const json j = json::parse(std::string(depth, '[') + '0' + std::string(depth, ']'));
+
+            json target;
+            target = j;
+
+            std::size_t target_depth = 0;
+            CHECK(*innermost_value(target, target_depth) == 0);
+            CHECK(target_depth == depth);
+        }
+
+        SECTION("depths around the bound of the recursive descent")
+        {
+            // The copy constructor descends into a bounded number of levels and
+            // completes whatever is below that without the call stack. Cover
+            // every depth around that bound, so that the two ways of copying
+            // are known to meet cleanly - wherever the bound is set.
+            for (std::size_t d = 1; d <= 300; ++d)
+            {
+                CAPTURE(d);
+
+                const json array = json::parse(std::string(d, '[') + '0' + std::string(d, ']'));
+                const json array_copy(array); // NOLINT(performance-unnecessary-copy-initialization): the copy is what is tested
+                std::size_t array_depth = 0;
+                CHECK(*innermost_value(array_copy, array_depth) == 0);
+                CHECK(array_depth == d);
+
+                std::string object_text;
+                for (std::size_t i = 0; i < d; ++i)
+                {
+                    object_text += "{\"a\":";
+                }
+                object_text += '1';
+                object_text.append(d, '}');
+
+                const json object = json::parse(object_text);
+                const json object_copy(object); // NOLINT(performance-unnecessary-copy-initialization): the copy is what is tested
+                std::size_t object_depth = 0;
+                CHECK(*innermost_value(object_copy, object_depth) == 1);
+                CHECK(object_depth == d);
+            }
+        }
+
+        SECTION("a value that is deep in one place only")
+        {
+            json j = json::object();
+            j["shallow"] = 1;
+            j["deep"] = json::parse(std::string(depth, '[') + '0' + std::string(depth, ']'));
+            j["also_shallow"] = json::array({1, 2, 3});
+
+            const json copy(j);
+
+            CHECK(copy["shallow"] == 1);
+            CHECK(copy["also_shallow"] == json::array({1, 2, 3}));
+
+            std::size_t deep_depth = 0;
+            CHECK(*innermost_value(copy["deep"], deep_depth) == 0);
+            CHECK(deep_depth == depth);
+        }
+
+        SECTION("comparing")
+        {
+            // Comparing used to descend once per level, and an ordered
+            // comparison used to compare every pair of elements twice, once in
+            // each direction, which took exponentially long in the nesting
+            // depth. Both are gone: these finish in milliseconds, where the
+            // second used to take longer than anyone would wait even for a
+            // value nested only a few dozen levels deep.
+            const std::string text = std::string(depth, '[') + '0' + std::string(depth, ']');
+            const json j = json::parse(text);
+            const json same = json::parse(text);
+            const json larger = json::parse(std::string(depth, '[') + '1' + std::string(depth, ']'));
+
+            CHECK(j == same);
+            CHECK_FALSE(j == larger);
+            CHECK(j != larger);
+
+            CHECK(j < larger);
+            CHECK_FALSE(larger < j);
+            CHECK(larger > j);
+            CHECK(j <= same);
+            CHECK(j >= same);
+
+            // a value that ends earlier is the smaller one
+            const json shorter = json::parse(std::string(depth - 1, '[') + '0' + std::string(depth - 1, ']'));
+            CHECK_FALSE(j == shorter);
+        }
+
+        SECTION("comparing objects")
+        {
+            std::string text;
+            text.reserve((6 * depth) + 1);
+            for (std::size_t i = 0; i < depth; ++i)
+            {
+                text += "{\"a\":";
+            }
+            text += '1';
+            text.append(depth, '}');
+
+            const json j = json::parse(text);
+            const json same = json::parse(text);
+
+            CHECK(j == same);
+            CHECK_FALSE(j != same);
+            CHECK(j <= same);
+            CHECK(j >= same);
+        }
+
+        SECTION("the copy is independent of the original")
+        {
+            const json j = json::parse(std::string(depth, '[') + '0' + std::string(depth, ']'));
+
+            json copy(j);
+
+            // reach the innermost value without recursing and replace it
+            json* current = &copy;
+            while (current->is_array() && !current->empty())
+            {
+                current = &current->front();
+            }
+            *current = 42;
+
+            std::size_t unused = 0;
+            CHECK(*innermost_value(copy, unused) == 42);
+            CHECK(*innermost_value(j, unused) == 0);
+        }
+    }
+}
+

tests/src/unit-merge_patch.cpp

@@ -14,6 +14,60 @@ using nlohmann::json;
     using namespace nlohmann::literals; // NOLINT(google-build-using-namespace)
 #endif
 
+#include <string>
+
+namespace
+{
+// RFC 7396's MergePatch, written recursively as in the RFC; only usable on
+// values nested a few hundred levels deep
+void reference_merge_patch(json& target, const json& patch)
+{
+    if (!patch.is_object())
+    {
+        target = patch;
+        return;
+    }
+    if (!target.is_object())
+    {
+        target = json::object();
+    }
+    for (auto it = patch.begin(); it != patch.end(); ++it)
+    {
+        if (it.value().is_null())
+        {
+            target.erase(it.key());
+        }
+        else
+        {
+            reference_merge_patch(target[it.key()], it.value());
+        }
+    }
+}
+
+// objects nested `depth` levels deep under the key "a", with members that
+// differ by `variant` on the way down
+std::string nested_objects(const std::size_t depth, const int variant)
+{
+    std::string text;
+    for (std::size_t i = 0; i < depth; ++i)
+    {
+        text += "{";
+        if ((i + static_cast<std::size_t>(variant)) % 3 == 0)
+        {
+            text += "\"s" + std::to_string(variant) + "\":" + std::to_string(i) + ",";
+        }
+        if (variant == 2 && i % 5 == 0)
+        {
+            text += "\"s0\":null,";
+        }
+        text += "\"a\":";
+    }
+    text += variant == 1 ? R"({"x":1,"y":null})" : "{\"y\":2}";
+    text.append(depth, '}');
+    return text;
+}
+} // namespace
+
 TEST_CASE("JSON Merge Patch")
 {
     SECTION("examples from RFC 7396")
@@ -242,3 +296,52 @@ TEST_CASE("JSON Merge Patch")
         }
     }
 }
+
+TEST_CASE("JSON Merge Patch on deeply nested values")
+{
+    SECTION("patching past the descent bound gives the same result")
+    {
+        // every depth on either side of where the iterative version takes
+        // over (detail::recursion_depth_limit(), 128)
+        for (std::size_t depth = 0; depth <= 300; ++depth)
+        {
+            CAPTURE(depth);
+            for (int variant = 0; variant < 3; ++variant)
+            {
+                CAPTURE(variant);
+                const json patch = json::parse(nested_objects(depth, variant));
+
+                json result = json::parse(nested_objects(depth, (variant + 1) % 3));
+                json expected = result;
+                result.merge_patch(patch);
+                reference_merge_patch(expected, patch);
+                CHECK(result == expected);
+
+                // a target that is not an object, and an empty one
+                json from_null;
+                from_null.merge_patch(patch);
+                json expected_from_null;
+                reference_merge_patch(expected_from_null, patch);
+                CHECK(from_null == expected_from_null);
+            }
+        }
+    }
+
+    SECTION("patches nested too deeply for the call stack (#5393)")
+    {
+        // applying a patch used to recurse once per nesting level. The result
+        // is only walked, never copied or compared, since those recurse too.
+        const std::size_t depth = 100000;
+        json target = json::parse(nested_objects(depth, 0));
+        target.merge_patch(json::parse(nested_objects(depth, 1)));
+
+        const json* p = &target;
+        for (std::size_t i = 0; i < depth; ++i)
+        {
+            p = &p->at("a");
+        }
+        // {"y":2} patched with {"x":1,"y":null}
+        CHECK(p->size() == 1);
+        CHECK(p->at("x") == 1);
+    }
+}

tests/src/unit-modifiers.cpp

@@ -11,6 +11,53 @@
 #include <nlohmann/json.hpp>
 using nlohmann::json;
 
+#include <string>
+
+namespace
+{
+// update(source, true) as documented, written recursively; only usable on
+// values nested a few hundred levels deep
+void reference_update(json& target, const json& source)
+{
+    for (auto it = source.begin(); it != source.end(); ++it)
+    {
+        const auto existing = target.find(it.key());
+        if (it.value().is_object() && existing != target.end() && existing->is_object())
+        {
+            reference_update(*existing, it.value());
+        }
+        else
+        {
+            target[it.key()] = it.value();
+        }
+    }
+}
+
+// objects nested `depth` levels deep under the key "a", with members that
+// differ by `variant` on the way down
+std::string nested_objects(const std::size_t depth, const int variant)
+{
+    std::string text;
+    for (std::size_t i = 0; i < depth; ++i)
+    {
+        text += "{";
+        if ((i + static_cast<std::size_t>(variant)) % 3 == 0)
+        {
+            text += "\"s" + std::to_string(variant) + "\":" + std::to_string(i) + ",";
+        }
+        if (variant == 2 && i % 5 == 0)
+        {
+            // an object replacing a primitive, which is not merged
+            text += R"("s0":{"o":1},)";
+        }
+        text += "\"a\":";
+    }
+    text += variant == 1 ? "{\"x\":1}" : "{\"y\":2}";
+    text.append(depth, '}');
+    return text;
+}
+} // namespace
+
 TEST_CASE("modifiers")
 {
     SECTION("clear()")
@@ -988,3 +1035,44 @@ TEST_CASE("modifiers")
         }
     }
 }
+
+TEST_CASE("update() on deeply nested values")
+{
+    SECTION("merging past the descent bound gives the same result")
+    {
+        // every depth on either side of where the iterative version takes
+        // over (detail::recursion_depth_limit(), 128)
+        for (std::size_t depth = 0; depth <= 300; ++depth)
+        {
+            CAPTURE(depth);
+            for (int variant = 0; variant < 3; ++variant)
+            {
+                CAPTURE(variant);
+                const json source = json::parse(nested_objects(depth, variant));
+                json result = json::parse(nested_objects(depth, (variant + 1) % 3));
+                json expected = result;
+                result.update(source, true);
+                reference_update(expected, source);
+                CHECK(result == expected);
+            }
+        }
+    }
+
+    SECTION("objects nested too deeply for the call stack (#5545)")
+    {
+        // merging used to recurse once per nesting level. The result is only
+        // walked, never copied or compared, since those recurse too.
+        const std::size_t depth = 100000;
+        json target = json::parse(nested_objects(depth, 0));
+        target.update(json::parse(nested_objects(depth, 1)), true);
+
+        const json* p = &target;
+        for (std::size_t i = 0; i < depth; ++i)
+        {
+            p = &p->at("a");
+        }
+        CHECK(p->size() == 2);
+        CHECK(p->at("x") == 1);
+        CHECK(p->at("y") == 2);
+    }
+}

tests/src/unit-ordered_json.cpp

@@ -82,6 +82,40 @@ TEST_CASE("regression test for issue #3732 - iteration_proxy_value<iter_impl<ord
     static_cast<void>(fn);
 }
 
+TEST_CASE("copying an ordered_json with nested values")
+{
+    // ordered_map is backed by a vector, so copying an object that has
+    // structured values takes a different route than copying a std::map-backed
+    // one; see https://github.com/nlohmann/json/issues/5387
+    ordered_json oj;
+    oj["z"] = 1;
+    oj["a"]["y"] = 2;
+    oj["a"]["b"]["x"] = 3;
+    oj["m"] = {1, 2, {{"w", 4}}};
+
+    const ordered_json copy(oj);
+
+    SECTION("the copy is equal to the original")
+    {
+        CHECK(copy == oj);
+        CHECK(copy.dump() == oj.dump());
+    }
+
+    SECTION("the key order is preserved at every level")
+    {
+        CHECK(copy.dump() == R"({"z":1,"a":{"y":2,"b":{"x":3}},"m":[1,2,{"w":4}]})");
+    }
+
+    SECTION("the copy is independent of the original")
+    {
+        ordered_json mutated(oj);
+        mutated["a"]["b"]["x"] = 99;
+
+        CHECK(oj["a"]["b"]["x"] == 3);
+        CHECK(mutated["a"]["b"]["x"] == 99);
+    }
+}
+
 TEST_CASE("regression test - diff() must account for ordered_json member order")
 {
     SECTION("pure reorder, no value changes")

tests/src/unit-precise-stream-position.cpp

@@ -0,0 +1,237 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  JSON for Modern C++ (supporting code)
+// |  |  |__   |  |  | | | |  version 3.12.0
+// |_____|_____|_____|_|___|  https://github.com/nlohmann/json
+//
+// SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>
+// SPDX-License-Identifier: MIT
+
+#include "doctest_compatibility.h"
+
+// This file tests the opt-in JSON_PRECISE_STREAM_POSITION, so it defines the
+// macro itself rather than relying on a -D flag, and runs in every build. The
+// default behavior is pinned in unit-deserialization.cpp.
+#ifdef JSON_PRECISE_STREAM_POSITION
+    #undef JSON_PRECISE_STREAM_POSITION
+#endif
+
+#define JSON_PRECISE_STREAM_POSITION 1
+
+#include <nlohmann/json.hpp>
+using nlohmann::json;
+
+#include <cstddef>
+#include <sstream>
+#include <streambuf>
+#include <string>
+#include <utility>
+#include <vector>
+
+#define STRINGIZE_EX(x) #x
+#define STRINGIZE(x) STRINGIZE_EX(x)
+
+namespace
+{
+// A streambuf that keeps no get area at all and refuses every putback: with an
+// empty get area, sungetc() always ends up in pbackfail(). Used to check that
+// the character terminating a number is left in the input without relying on
+// the streambuf being able to put a consumed character back.
+class no_putback_streambuf : public std::streambuf
+{
+  public:
+    explicit no_putback_streambuf(std::string s) : m_data(std::move(s)) {}
+
+  protected:
+    // peek at the next character without consuming it
+    int_type underflow() override
+    {
+        if (m_pos >= m_data.size())
+        {
+            return traits_type::eof();
+        }
+        return traits_type::to_int_type(m_data[m_pos]);
+    }
+
+    // consume the next character
+    int_type uflow() override
+    {
+        if (m_pos >= m_data.size())
+        {
+            return traits_type::eof();
+        }
+        return traits_type::to_int_type(m_data[m_pos++]);
+    }
+
+    int_type pbackfail(int_type /*c*/) override
+    {
+        return traits_type::eof();
+    }
+
+  private:
+    std::string m_data;
+    std::size_t m_pos = 0;
+};
+
+// read the characters that are left in a stream
+std::string remaining(std::istream& is)
+{
+    std::string result;
+    char c = 0;
+    while (is.get(c))
+    {
+        result += c;
+    }
+    return result;
+}
+}  // namespace
+
+TEST_CASE("JSON_PRECISE_STREAM_POSITION")
+{
+    SECTION("the macro is part of the ABI tag")
+    {
+        const std::string ns = STRINGIZE(NLOHMANN_JSON_NAMESPACE);
+        // other tags may come before it, e.g. json_abi_diag_psp
+        CHECK(ns.find("_psp") != std::string::npos);
+    }
+
+    SECTION("a number does not consume the character that terminates it")
+    {
+        // a number is only terminated by the character following it; that
+        // character must be given back so the stream is positioned right
+        // after the value
+        const std::vector<std::pair<std::string, std::string>> tests =
+        {
+            {"1true", "true"},
+            {"1[2]", "[2]"},
+            {"1{}", "{}"},
+            {R"(1"a")", R"("a")"},
+            {"1 true", " true"},
+            {"12,", ","},
+            {"-0.5e3x", "x"},
+            {"1null", "null"}
+        };
+
+        for (const auto& test : tests)
+        {
+            CAPTURE(test.first);
+            std::istringstream ss(test.first);
+            json j;
+            ss >> j;
+            CHECK(j == json::parse(test.first.substr(0, test.first.size() - test.second.size())));
+            CHECK(remaining(ss) == test.second);
+        }
+    }
+
+    SECTION("values that are self-delimiting are unaffected")
+    {
+        const std::vector<std::pair<std::string, std::string>> tests =
+        {
+            {"truefalse", "false"},
+            {"[1][2]", "[2]"},
+            {R"({"a":1}{"b":2})", R"({"b":2})"},
+            {R"("a""b")", R"("b")"},
+            {"null null", " null"}
+        };
+
+        for (const auto& test : tests)
+        {
+            CAPTURE(test.first);
+            std::istringstream ss(test.first);
+            json j;
+            ss >> j;
+            CHECK(remaining(ss) == test.second);
+        }
+    }
+
+    SECTION("a number at the end of the input leaves nothing behind")
+    {
+        for (const std::string s :
+                {"1", "12", "-3.5e2", " 7 "
+                })
+        {
+            CAPTURE(s);
+            std::istringstream ss(s);
+            json j;
+            ss >> j;
+            CHECK(remaining(ss).find_first_not_of(" \t\n\r") == std::string::npos);
+        }
+    }
+
+    SECTION("repeated extraction of concatenated values")
+    {
+        std::istringstream ss(R"(1true[2]3"x"{"a":4}5)");
+        const std::vector<json> expected =
+        {
+            json(1), json(true), json::parse("[2]"), json(3),
+            json("x"), json::parse(R"({"a":4})"), json(5)
+        };
+
+        for (const auto& e : expected)
+        {
+            json j;
+            ss >> j;
+            CHECK(j == e);
+        }
+    }
+
+    SECTION("differences to the default behavior")
+    {
+        // both of these work by accident without the macro, because the
+        // character after a number is swallowed; see unit-deserialization.cpp
+
+        SECTION("a separator after a number is not skipped")
+        {
+            std::istringstream ss("1,2");
+            json j;
+            ss >> j;
+            CHECK(j == 1);
+            CHECK_THROWS_AS(ss >> j, json::parse_error&);
+        }
+
+        SECTION("std::getline after a number sees the line break")
+        {
+            std::istringstream ss("42\nfoo");
+            json j;
+            std::string line;
+            ss >> j;
+            std::getline(ss, line);
+            CHECK(j == 42);
+            CHECK(line.empty());
+            std::getline(ss, line);
+            CHECK(line == "foo");
+        }
+    }
+
+    SECTION("sax_parse with strict == false")
+    {
+        std::istringstream ss("1true");
+        json j;
+        nlohmann::detail::json_sax_dom_parser<json, nlohmann::detail::input_stream_adapter> sdp(j, true);
+        CHECK(json::sax_parse(ss, &sdp, nlohmann::detail::input_format_t::json, false));
+        CHECK(j == 1);
+        CHECK(remaining(ss) == "true");
+    }
+
+    SECTION("strict parsing still rejects trailing data")
+    {
+        std::istringstream ss("1true");
+        json _;
+        CHECK_THROWS_WITH_AS(_ = json::parse(ss),
+                             "[json.exception.parse_error.101] parse error at line 1, column 5: syntax error while parsing value - unexpected true literal; expected end of input", json::parse_error&);
+
+        std::istringstream ss2("1true");
+        CHECK_FALSE(json::accept(ss2));
+    }
+
+    SECTION("a streambuf that cannot put back is not needed")
+    {
+        // the terminating character is never consumed, so no putback
+        // position is required
+        no_putback_streambuf buf("1true");
+        std::istream is(&buf);
+        json j;
+        is >> j;
+        CHECK(j == json(1));
+        CHECK(remaining(is) == "true");
+    }
+}

tests/src/unit-regression3.cpp

@@ -658,33 +658,6 @@ TEST_CASE("regression test #5074 - portable workaround for single-element brace
     CHECK(j[0] == j_obj);
 }
 
-#if defined(JSON_BRACE_INIT_COPY_SEMANTICS) && (JSON_BRACE_INIT_COPY_SEMANTICS == 1)
-TEST_CASE("regression test #5074 - single-element brace init with JSON_BRACE_INIT_COPY_SEMANTICS")
-{
-    // with JSON_BRACE_INIT_COPY_SEMANTICS: single-element brace init copies/moves
-    json const j_obj = {{"key", "value"}, {"num", 42}};
-    json const j_arr = {1, 2, 3};
-
-    // object: brace init copies instead of wrapping
-    json const j1{j_obj};
-    CHECK(j1.is_object());
-    CHECK(j1 == j_obj);
-
-    // array: brace init copies instead of wrapping
-    json const j2{j_arr};
-    CHECK(j2.is_array());
-    CHECK(j2.size() == 3);
-    CHECK(j2 == j_arr);
-
-    // primitives still work as initializer lists
-    json const j3{true};
-    CHECK(j3.is_boolean());
-
-    json const j4{42};
-    CHECK(j4.is_number_integer());
-}
-#endif
-
 struct Example_5122
 {
     float b = 2;

tests/src/unit-ubjson.cpp

@@ -15,6 +15,7 @@ using nlohmann::json;
 #include <fstream>
 #include <set>
 #include "make_test_data_available.hpp"
+#include "round_trip_corpus.hpp"
 #include "test_utils.hpp"
 
 namespace
@@ -2265,7 +2266,9 @@ TEST_CASE("UBJSON optimized arrays of a valueless type are bounded")
 
     SECTION("an excessive count is rejected")
     {
-        // 'l' is a big-endian int32: 0x7FFFFFFF elements, about 34 GB of value
+        // 'l' is a big-endian int32: 0x7FFFFFFF elements, about 34 GB of value;
+        // OSS-Fuzz reported this shape as a parse_ubjson_fuzzer timeout
+        // (testcase 6347769435193344, no issue filed)
         for (const auto marker :
                 {'Z', 'T', 'F'
                 })
@@ -2817,6 +2820,51 @@ TEST_CASE("UBJSON use_type requires use_size")
     }
 }
 
+TEST_CASE("UBJSON round-trip invariants")
+{
+    // This checks what the parse_ubjson_fuzzer driver checks (see
+    // tests/src/fuzzer-parse_ubjson.cpp), so that a regression shows up in CI
+    // rather than as an OSS-Fuzz report: every value from_ubjson() returns
+    // (j1) can be serialized with any combination of options, the result can
+    // be parsed back (j2), and serializing j2 again with the same options
+    // reproduces the exact bytes. Beyond the driver, this also checks that j2
+    // equals j1. Values are compared with dump() rather than operator==,
+    // because a NaN never compares equal to itself.
+    struct options
+    {
+        bool use_size;
+        bool use_type;
+    };
+    const std::vector<options> all_options =
+    {
+        {false, false},
+        {true, false},
+        {true, true},
+    };
+
+    for (const auto& j0 : utils::round_trip_corpus::values())
+    {
+        // turn the corpus value into a value as from_ubjson() returns it; this
+        // has no binary values, as UBJSON writes them as arrays of integers
+        for (const auto& initial : all_options)
+        {
+            const json j1 = json::from_ubjson(json::to_ubjson(j0, initial.use_size, initial.use_type));
+
+            for (const auto& o : all_options)
+            {
+                INFO("j1 = " << j1.dump() << ", use_size = " << o.use_size << ", use_type = " << o.use_type);
+
+                const std::vector<std::uint8_t> vec = json::to_ubjson(j1, o.use_size, o.use_type);
+                json j2;
+                // anything the library writes must be parsable by the library
+                REQUIRE_NOTHROW(j2 = json::from_ubjson(vec));
+                CHECK(j2.dump() == j1.dump());
+                CHECK(json::to_ubjson(j2, o.use_size, o.use_type) == vec);
+            }
+        }
+    }
+}
+
 TEST_CASE("UBJSON roundtrips" * doctest::skip())
 {
     SECTION("input from self-generated UBJSON files")

tools/generate_natvis/generate_natvis.py

@@ -20,7 +20,7 @@ if __name__ == '__main__':
 
     namespaces = ['nlohmann']
     abi_prefix = 'json_abi'
-    abi_tags = ['_diag', '_ldvcmp']
+    abi_tags = ['_diag', '_ldvcmp', '_dp', '_bics', '_psp']
     version = '_v' + args.version.replace('.', '_')
     inline_namespaces = []
 

tools/serve_header/README.md

@@ -60,6 +60,10 @@ int main() {
 `serve_header.py` will try to read a configuration file `serve_header.yml` in the top level or project root directory, and will fall back on built-in defaults if the file cannot be read.
 An annotated example configuration can be found in `tools/serve_header/serve_header.yml.example`.
 
+By default, the server listens on `localhost` only, and only web pages from Compiler Explorer (`https://godbolt.org` and `https://compiler-explorer.com`) may read the header.
+Set `bind` to serve other machines as well; anyone who can reach the server can then trigger `make` runs in your working trees.
+Set `cors_origins` to allow other web pages.
+
 ## Serving `json.hpp` from multiple project directory instances or working trees
 
 `serve_header.py` was designed with the goal of supporting multiple project roots or working trees at the same time.

tools/serve_header/serve_header.py

@@ -26,6 +26,10 @@ HEADER = 'json.hpp'
 
 DATETIME_FORMAT = '%Y-%m-%d %H:%M:%S'
 
+# origins whose pages may read the served header from a browser; Compiler
+# Explorer downloads #include <https://...> headers client-side
+DEFAULT_CORS_ORIGINS = ['https://godbolt.org', 'https://compiler-explorer.com']
+
 JSON_VERSION_RE = re.compile(r'\s*#\s*define\s+NLOHMANN_JSON_VERSION_MAJOR\s+')
 
 class ExitHandler(logging.StreamHandler):
@@ -247,6 +251,8 @@ class WorkTrees(FileSystemEventHandler):
         self.observer.join()
 
 class HeaderRequestHandler(SimpleHTTPRequestHandler): # lgtm[py/missing-call-to-init]
+    cors_origins = DEFAULT_CORS_ORIGINS
+
     def __init__(self, request, client_address, server):
         """."""
         self.worktrees = server.worktrees
@@ -310,8 +316,11 @@ class HeaderRequestHandler(SimpleHTTPRequestHandler): # lgtm[py/missing-call-to-
 
         # set content length
         super().send_header('Content-Length', length)
-        # CORS header
-        self.send_header('Access-Control-Allow-Origin', '*')
+        # CORS header; only for the configured origins
+        origin = self.headers.get('Origin')
+        if origin in self.cors_origins:
+            self.send_header('Access-Control-Allow-Origin', origin)
+        self.send_header('Vary', 'Origin')
         # prevent caching
         self.send_header('Cache-Control', 'no-cache, no-store, must-revalidate')
         self.send_header('Pragma', 'no-cache')
@@ -383,8 +392,15 @@ if __name__ == '__main__':
         # find and monitor working trees
         worktrees = WorkTrees(config.get('root', '.'))
 
-        # start web server
-        infos = socket.getaddrinfo(config.get('bind', None), config.get('port', 8443),
+        # origins allowed to read the header from a browser
+        cors_origins = config.get('cors_origins', DEFAULT_CORS_ORIGINS)
+        if isinstance(cors_origins, str):
+            cors_origins = [cors_origins]
+        HeaderRequestHandler.cors_origins = cors_origins
+
+        # start web server; only reachable from this machine unless configured
+        # otherwise (bind: null listens on all interfaces)
+        infos = socket.getaddrinfo(config.get('bind', 'localhost'), config.get('port', 8443),
                                    type=socket.SOCK_STREAM, flags=socket.AI_PASSIVE)
         DualStackServer.address_family = infos[0][0]
         HeaderRequestHandler.protocol_version = 'HTTP/1.0'

tools/serve_header/serve_header.yml.example

@@ -10,6 +10,13 @@
 #  cert_file: localhost.pem
 #  key_file: localhost-key.pem
 
-# address and port for the server to listen on
-# bind: null
+# address and port for the server to listen on; by default, only this machine
+# can connect. Binding to a network address, or to null for all interfaces,
+# lets other machines connect, and every request runs make in a working tree.
+# bind: localhost
 # port: 8443
+
+# origins whose web pages may read the header (CORS)
+# cors_origins:
+#   - https://godbolt.org
+#   - https://compiler-explorer.com