Bound UBJSON optimized arrays of a valueless type

An element of type 'Z' (null), 'T' (true) or 'F' (false) is encoded by its type marker alone, so an optimized UBJSON array of one of those has no payload: reading an element consumes no input at all. Its declared count is therefore the only thing that decides how much is allocated, and nothing bounded it. "[$Z#l" and a four-byte count is nine bytes of input describing two billion values; #2793 reports 35 GB and 150 seconds from ten bytes, and OSS-Fuzz has an out-of-memory and a timeout report for the same shape. Every other type costs at least one byte per element, so the end of the input bounds it. 'N' (no-op) is already skipped rather than stored. Objects are not affected either: each element is preceded by its key, which costs bytes. And BJData already refuses these markers as an optimized type, so this is a plain UBJSON matter. Reject a count above 1,048,576 elements for those three types with out_of_range.408, the code this reader already uses for a declared size it will not honour. The check runs before the SAX start event, so no container is opened and then abandoned. Rejecting on the read side alone would break the guarantee that anything to_ubjson() writes can be read back, and would trip the round-trip assertion in fuzzer-parse_ubjson.cpp. So the writer falls back to the unoptimized encoding, one byte per element, for arrays of these types above the same limit. Its decision depends only on the array's size, which is identical for a value and for anything parsed back from it, so the round trip is stable. No existing test changes: the largest such count in the test suite is 65,793. The excessive-size test that already used this shape still passes, now rejected a little earlier than by the max_size() check it used to reach. Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Reject a nested BJData ndarray dimension vector where it is read
2026-09-07 00:37:58 +00:00 · 2026-09-06 18:12:41 +02:00 · 2026-09-06 18:12:40 +02:00 · 2026-09-06 18:12:39 +02:00 · 2026-09-06 16:41:35 +02:00 · 2026-09-05 20:30:22 +02:00
19 changed files with 1018 additions and 189 deletions
@@ -38,14 +38,14 @@ jobs:

    # Initializes the CodeQL tools for scanning.
    - name: Initialize CodeQL
-      uses: github/codeql-action/init@db488ddef3bf6cb639b32c2e9a7c0a7ea8271d28 # v4.37.8
+      uses: github/codeql-action/init@cdf488f595d80d6e07e03d4674febd5ab45fa938 # v4.37.9
      with:
        languages: c-cpp

    # Autobuild attempts to build any compiled languages  (C/C++, C#, or Java).
    # If this step fails, then you should remove it and run the build manually (see below)
    - name: Autobuild
-      uses: github/codeql-action/autobuild@db488ddef3bf6cb639b32c2e9a7c0a7ea8271d28 # v4.37.8
+      uses: github/codeql-action/autobuild@cdf488f595d80d6e07e03d4674febd5ab45fa938 # v4.37.9

    - name: Perform CodeQL Analysis
-      uses: github/codeql-action/analyze@db488ddef3bf6cb639b32c2e9a7c0a7ea8271d28 # v4.37.8
+      uses: github/codeql-action/analyze@cdf488f595d80d6e07e03d4674febd5ab45fa938 # v4.37.9
@@ -43,6 +43,6 @@ jobs:
          output: 'flawfinder_results.sarif'

      - name: Upload analysis results to GitHub Security tab
-        uses: github/codeql-action/upload-sarif@db488ddef3bf6cb639b32c2e9a7c0a7ea8271d28 # v4.37.8
+        uses: github/codeql-action/upload-sarif@cdf488f595d80d6e07e03d4674febd5ab45fa938 # v4.37.9
        with:
          sarif_file: ${{github.workspace}}/flawfinder_results.sarif
@@ -76,6 +76,6 @@ jobs:

      # Upload the results to GitHub's code scanning dashboard.
      - name: "Upload to code-scanning"
-        uses: github/codeql-action/upload-sarif@db488ddef3bf6cb639b32c2e9a7c0a7ea8271d28 # v4.37.8
+        uses: github/codeql-action/upload-sarif@cdf488f595d80d6e07e03d4674febd5ab45fa938 # v4.37.9
        with:
          sarif_file: results.sarif
@@ -61,7 +61,7 @@ jobs:

      # Upload SARIF file generated in previous step
      - name: Upload SARIF file
-        uses: github/codeql-action/upload-sarif@db488ddef3bf6cb639b32c2e9a7c0a7ea8271d28 # v4.37.8
+        uses: github/codeql-action/upload-sarif@cdf488f595d80d6e07e03d4674febd5ab45fa938 # v4.37.9
        with:
          sarif_file: semgrep.sarif
        if: always()
@@ -125,6 +125,7 @@ The library uses the following mapping from JSON values types to BJData types ac

    - `"_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
    - every element of `"_ArrayData_"` is a number of the kind named by `"_ArrayType_"` (a floating-point number for
@@ -69,6 +69,12 @@ The library uses the following mapping from JSON values types to UBJSON types ac
    Note that `use_size = true` alone may result in larger representations - the benefit of this parameter is that the
    receiving side is immediately informed on the number of elements of the container.

+    An array whose type marker is `Z` (null), `T` (true) or `F` (false) stores no payload at all, because the marker
+    already is the value. Its declared count is therefore the only thing that decides how much memory the receiving side
+    allocates, and a handful of bytes can describe billions of elements. `from_ubjson` rejects such an array with
+    [`out_of_range.408`](../../home/exceptions.md#jsonexceptionout_of_range408) when the count exceeds 1,048,576, and
+    `to_ubjson` writes longer arrays of these types without the annotation, so any value it produces can be read back.
+
 !!! info "Binary values"

    If the JSON data contains the binary type, the value stored is a list of integers, as suggested by the UBJSON
@@ -8,41 +8,72 @@ the result of an internet search. If you know further customers of the library,
 ## Space Exploration

 - [**Peregrine Lunar Lander Flight 01**](https://en.wikipedia.org/wiki/Peregrine_Mission_One) - The library was used for payload management in the **Peregrine Moon Lander**, developed by **Astrobotic Technology** and launched as part of NASA's **Commercial Lunar Payload Services (CLPS)** program. After six days in orbit, the spacecraft was intentionally redirected into Earth's atmosphere, where it burned up over the Pacific Ocean on **January 18, 2024**.
+- [**NASA Unsteady Pressure-Sensitive Paint Processing**](https://github.com/nasa/upsp-processing), NASA software for processing high-speed video recordings of wind tunnel tests on launch vehicle and aircraft models
+- [**Terma TEMU**](https://temu.terma.com/docs/public/temu-release-notes/latest/copying/json-for-modern-cpp.html), an emulator of spacecraft on-board computers used to develop and validate flight software for European space missions

 ## Automotive

 - [**Alexa Auto SDK**](https://github.com/alexa/alexa-auto-sdk), a software development kit enabling the integration of Alexa into automotive systems
 - [**Apollo**](https://github.com/ApolloAuto/apollo), a framework for building autonomous driving systems
 - [**Automotive Grade Linux (AGL)**](https://download.automotivelinux.org/AGL/release/jellyfish/latest/qemux86-64/deploy/licenses/nlohmann-json/), a collaborative open-source platform for automotive software development
+- [**Autoware**](https://github.com/autowarefoundation/autoware_universe), an open-source software stack for autonomous driving built on ROS 2
+- [**Eclipse S-CORE**](https://github.com/eclipse-score/nlohmann_json), an open-source software platform for the software-defined vehicle backed by major automotive manufacturers and suppliers
 - [**Genesis Motor** (infotainment)](http://webmanual.genesis.com/ccIC/AVNT/JW/KOR/English/reference010.html), a luxury automotive brand
 - [**Hyundai** (infotainment)](https://www.hyundai.com/wsvc/ww/download.file.do?id=/content/hyundai/ww/data/opensource/data/GN7-2022/licenseCode/info), a global automotive brand
 - [**Kia** (infotainment)](http://webmanual.kia.com/PREM_GEN6/AVNT/RJPE/KOR/Korean/reference010.html), a global automotive brand
 - [**Mercedes-Benz Operating System (MB.OS)**](https://group.mercedes-benz.com/careers/about-us/mercedes-benz-operating-system/), a core component of the vehicle software ecosystem from Mercedes-Benz
+- [**NVIDIA DRIVE OS**](https://developer.nvidia.com/docs/drive/drive-os/6.0.5/public/driveworks-nvcgf/dwx_open_source_attribution.html), the operating system and DriveWorks SDK powering NVIDIA's platform for autonomous vehicles
 - [**Rivian** (infotainment)](https://assets.ctfassets.net/2md5qhoeajym/3cwyo4eoufk4yingUwusFt/ded2c47da620fdfc99c88c7156d2c1d8/In-Vehicle_OSS_Attribution_2024__11-24_.pdf), an electric vehicle manufacturer
 - [**Suzuki** (infotainment)](https://www.globalsuzuki.com/motorcycle/ipc/oss/oss_48KA_00.pdf), a global automotive and motorcycle manufacturer

 ## Gaming and Entertainment

+- [**Anno 117: Pax Romana**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a city-building strategy game set in the Roman Empire
 - [**Assassin's Creed: Mirage**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a stealth-action game set in the Middle East, focusing on the journey of a young assassin with classic parkour and stealth mechanics
+- [**Battlefield 6**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a military first-person shooter known for its large-scale multiplayer battles
+- [**Battlefield: REDSEC**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a free-to-play battle royale experience set in the Battlefield universe
+- [**BioMenace: Remastered**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a remaster of the classic side-scrolling platform shooter
 - [**Chasm: The Rift**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a first-person shooter blending horror and adventure, where players navigate dark realms and battle monsters
 - [**College Football 25**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a college football simulation game featuring gameplay that mimics real-life college teams and competitions
+- [**College Football 26**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a college football simulation game featuring licensed teams and stadiums
+- [**College Football 27**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), the latest installment of the college football simulation series
 - [**Concepts**](https://concepts.app/en/licenses), a digital sketching app designed for creative professionals, offering flexible drawing tools for illustration, design, and brainstorming
 - [**Depthkit**](https://www.depthkit.tv/third-party-licenses), a tool for creating and capturing volumetric video, enabling immersive 3D experiences and interactive content
+- [**Dune: Awakening**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), an open-world survival MMO set on the desert planet Arrakis
+- [**EA Sports FC 25**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), an association football simulation with club, career, and online modes
+- [**EA Sports FC 26**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), the latest installment of the association football simulation series
+- [**EA Sports UFC 6**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a mixed martial arts fighting simulation
+- [**FiveM**](https://github.com/citizenfx/fivem), a modification framework for Grand Theft Auto V that powers custom multiplayer servers
+- [**FLUX:: Immersive**](https://doc.flux.audio/syrah/Credits.html), a suite of professional audio processing and immersive mixing plugins used in music and post-production
 - [**IMG.LY**](https://img.ly/acknowledgements), a platform offering creative tools and SDKs for integrating advanced image and video editing in applications
+- [**immersivetech**](https://immersitech.io/open-source-third-party-software/), a technology company focused on immersive experiences, providing tools and solutions for virtual and augmented reality applications
+- [**Kodi**](https://github.com/xbmc/xbmc/blob/master/xbmc/utils/JSONVariantWriter.cpp), a home theater and media center application
 - [**LOOT**](https://loot.readthedocs.io/_/downloads/en/0.13.0/pdf/), a tool for optimizing the load order of game plugins, commonly used in The Elder Scrolls and Fallout series
+- [**LunaTranslator**](https://github.com/HIllya51/LunaTranslator/blob/main/src/NativeImpl/LunaSubprocess/aspatch.cpp), a real-time translation tool for visual novels
+- [**MaaAssistantArknights**](https://github.com/MaaAssistantArknights/MaaAssistantArknights/blob/dev-v2/src/MaaCore/Vision/Roguelike/BlackFlow/BlackFlowMapAnalyzer.cpp), an automation assistant for the mobile game Arknights
 - [**Madden NFL 25**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a sports simulation game capturing the excitement of American football with realistic gameplay and team management features
+- [**Madden NFL 26**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), an American football simulation with franchise and team management modes
+- [**Madden NFL 27**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), the latest installment of the American football simulation series
 - [**Marne**](https://marne.io/licenses), an unofficial private server platform for hosting custom Battlefield 1 game experiences
 - [**Minecraft**](https://www.minecraft.net/zh-hant/attribution), a popular sandbox video game
+- [**Mumble**](https://github.com/mumble-voip/mumble), a low-latency, open-source voice chat application widely used by gaming communities
 - [**NHL 22**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a hockey simulation game offering realistic gameplay, team management, and various modes to enhance the hockey experience
+- [**OBS Studio**](https://github.com/obsproject/obs-studio), a free and open-source suite for video recording and live streaming
+- [**OpenRCT2**](https://github.com/OpenRCT2/OpenRCT2/blob/develop/src/openrct2/core/JsonFwd.hpp), an open source re-implementation of RollerCoaster Tycoon 2
 - [**Pixelpart**](https://pixelpart.net/documentation/book/third-party.html), a 2D animation and video compositing software that allows users to create animated graphics and visual effects with a focus on simplicity and ease of use
 - [**Razer Cortex**](https://mysupport.razer.com/app/answers/detail/a_id/14146/~/open-source-software-for-razer-software), a gaming performance optimizer and system booster designed to enhance the gaming experience
 - [**Red Dead Redemption II**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), an open-world action-adventure game following an outlaw's story in the late 1800s, emphasizing deep storytelling and immersive gameplay
+- [**RetroArch**](https://github.com/libretro/RetroArch), a frontend for emulators, game engines, and media players built on the libretro API
+- [**shadPS4**](https://github.com/shadps4-emu/shadPS4/blob/main/src/core/user_manager.h), a PlayStation 4 emulator for Windows, Linux and macOS
+- [**skate.**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a free-to-play skateboarding game set in an open world
 - [**Snapchat**](https://www.snap.com/terms/license-android), a multimedia messaging and augmented reality app for communication and entertainment
+- [**Steel Century Groove**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), an action game released in 2026
+- [**Sunshine**](https://github.com/LizardByte/Sunshine/blob/master/src/confighttp.h), a self-hosted game streaming host compatible with Moonlight clients
 - [**Tactics Ogre: Reborn**](https://www.square-enix-games.com/en_US/documents/tactics-ogre-reborn-pc-installer-software-and-associated-plug-ins-disclosure), a tactical role-playing game featuring strategic battles and deep storytelling elements
 - [**Throne and Liberty**](https://www.amazon.com/gp/help/customer/display.html?nodeId=T7fLNw5oAevCMtJFPj&pop-up=1), an MMORPG that offers an expansive fantasy world with dynamic gameplay and immersive storytelling
 - [**Unity Vivox**](https://docs.unity3d.com/Packages/com.unity.services.vivox@15.1/license/Third%20Party%20Notices.html), a communication service that enables voice and text chat functionality in multiplayer games developed with Unity
+- [**xemu**](https://github.com/xemu-project/xemu), an emulator of the original Xbox console
 - [**Zool: Redimensioned**](https://www.mobygames.com/person/1195889/niels-lohmann/credits/), a modern reimagining of the classic platformer featuring fast-paced gameplay and vibrant environments
- [**immersivetech**](https://immersitech.io/open-source-third-party-software/), a technology company focused on immersive experiences, providing tools and solutions for virtual and augmented reality applications

 ## Consumer Electronics

@@ -50,109 +81,195 @@ the result of an internet search. If you know further customers of the library,
 - [**Canon CanoScan LIDE**](https://carolburo.com/wp-content/uploads/2024/06/LiDE400_OnlineManual_Win_FR_V02.pdf), a series of flatbed scanners offering high-resolution image scanning for home and office use
 - [**Canon PIXMA Printers**](https://www.mediaexpert.pl/products/files/73/7338196/Instrukcja-obslugi-CANON-Pixma-TS7450i.pdf), a line of all-in-one inkjet printers known for high-quality printing and wireless connectivity
 - [**Cisco Webex Desk Camera**](https://www.cisco.com/c/dam/en_us/about/doing_business/open_source/docs/CiscoWebexDeskCamera-23-1622100417.pdf), a video camera designed for professional-quality video conferencing and remote collaboration
+- [**DJI Edge SDK**](https://github.com/dji-sdk/Edge-SDK-V2-Demo), the reference applications for DJI's Edge SDK, used to build edge computing services on DJI drone docks
+- [**Elgato Stream Deck**](https://github.com/elgatosf/streamdeck-obs-plugin2), a family of programmable control surfaces for content creators and their plugin ecosystem
+- [**Instagrid**](https://instagrid.co/intellectual-property/foss), a manufacturer of portable, high-performance battery systems for professional mobile power supply
+- [**iRobot**](https://iot-content.irobot.com/iw/sfsites/c/cms/delivery/media/MCKRLTPDJSSJBNJKDA5SG5UVVIIQ), a manufacturer of autonomous home robots including the Roomba vacuum cleaner range
+- [**Logitech Logi Bolt**](https://opensource.logitech.com/wiki/Logi_BoltApp/), the management application for Logitech's secure wireless connectivity technology
+- [**Novitus**](https://novitus.pl/licencjepensource), a manufacturer of fiscal cash registers and point-of-sale devices
 - [**Philips Hue Personal Wireless Lighting**](http://2ak5ape.257.cz/), a smart lighting system for customizable and wireless home illumination
 - [**Ray-Ban Meta Smart glasses**](https://www.meta.com/de/en/legal/smart-glasses/third-party-notices-android/03/), a pair of smart glasses designed for capturing photos and videos with integrated connectivity and social features
 - [**Razer Synapse**](https://mysupport.razer.com/app/answers/detail/a_id/14146/~/open-source-software-for-razer-software), a unified configuration software enabling hardware customization for Razer devices
+- [**Sharp Professional Displays**](https://jp.sharp/restricted/business/lcd-display/cms/images/source_pnla862/PN-LA652_752_862_LicenseInformation.pdf), a range of large-format interactive displays for business and education
 - [**Siemens SINEMA Remote Connect**](https://cache.industry.siemens.com/dl/files/790/109793790/att_1054961/v2/OSS_SINEMA-RC_86.pdf), a remote connectivity solution for monitoring and managing industrial networks and devices securely
+- [**Skydio**](https://pages.skydio.com/rs/784-TUF-591/images/Open%20Source%20Software%20Notice%20v0.2.html), a manufacturer of autonomous drones for inspection, public safety, and defense applications
 - [**Sony PlayStation 4**](https://doc.dl.playstation.net/doc/ps4-oss/index.html), a gaming console developed by Sony that offers a wide range of games and multimedia entertainment features
+- [**Sony Spatial Reality Display**](https://www.sony.co.jp/en/Products/Developer-Spatial-Reality-display/download/dcc-tools/blender-plugin/SpatiaRealityDisplayPluginforPreviewBL_Manual.pdf), a glasses-free stereoscopic 3D display and its plugins for Blender, 3ds Max, and ZBrush
 - [**Sony Virtual Webcam Driver for Remote Camera**](https://helpguide.sony.net/rc/vwd/v1/zh-cn/print.pdf), a software driver that enables the use of Sony cameras as virtual webcams for video conferencing and streaming
+- [**Yamaha Clavinova**](https://usa.yamaha.com/files/download/other_assets/1/2298171/CLP-800_oss_license.pdf), a series of digital pianos combining acoustic piano feel with digital sound technology

-## Operating Systems
+## Operating Systems and Platforms

 - [**Apple iOS and macOS**](https://www.apple.com/macos), a family of operating systems developed by Apple, including iOS for mobile devices and macOS for desktop computers
+- [**Chromium**](https://chromium.googlesource.com/chromium/src/+/main/third_party/nlohmann_json/), the open-source browser project that Google Chrome, Microsoft Edge, and many other browsers are built on, where the library is used as data container for on-device model execution
 - [**Google Fuchsia**](https://fuchsia.googlesource.com/third_party/json/), an open-source operating system developed by Google, designed to be secure, updatable, and adaptable across various devices
+- [**LG webOS**](https://github.com/webosose/com.webos.service.camera), a Linux-based operating system used in LG smart TVs, signage, and embedded devices
+- [**Microsoft Azure Linux**](https://github.com/microsoft/azurelinux), a Linux distribution developed by Microsoft for Azure infrastructure and edge workloads
+- [**OpenHarmony**](https://github.com/openharmony/third_party_json), an open-source operating system for smart devices and the foundation of HarmonyOS
 - [**SerenityOS**](https://github.com/SerenityOS/serenity), an open-source operating system that aims to provide a simple and beautiful user experience with a focus on simplicity and elegance
+- [**Windows Subsystem for Linux**](https://github.com/microsoft/WSL), a compatibility layer that runs Linux environments natively on Windows
 - [**Yocto**](http://ftp.emacinc.com/openembedded-sw/kirkstone-icop-5.15-kirkstone-6.0/archive-2024-10/pn8m-090t-ppc/licenses/nlohmann-json/), a Linux-based build system for creating custom operating systems and software distributions, tailored for embedded devices and IoT applications

 ## Development Tools and IDEs

 - [**Accentize SpectralBalance**](https://www.accentize.com/products/SpectralBalanceManual.pdf), an adaptive speech analysis tool designed to enhance audio quality by optimizing frequency balance in recordings
+- [**Airbus Ghidralligator**](https://www.cyber.airbus.com/en/newsroom/stories/2025-06-ghidralligator), a Ghidra-based emulator from Airbus CyberSecurity used to fuzz and analyse embedded firmware
+- [**Apache brpc**](https://github.com/apache/brpc/blob/master/src/butil/iobuf.h), an industrial-grade remote procedure call framework for C++
 - [**Arm Compiler for Linux**](https://documentation-service.arm.com/static/66558e9d876c8d213b7843e4), a software development toolchain for compiling and optimizing applications on Arm-based Linux systems
 - [**BBEdit**](https://s3.amazonaws.com/BBSW-download/BBEdit_15.1.2_User_Manual.pdf), a professional text and code editor for macOS
 - [**CoderPad**](https://coderpad.io), a collaborative coding platform that enables real-time code interviews and assessments for developers; the library is included in every CoderPad instance and can be accessed with a simple `#include "json.hpp"`
+- [**Codon**](https://github.com/exaloop/codon/blob/develop/jupyter/jupyter.h), an ahead-of-time compiler for a Python-like language
 - [**Compiler Explorer**](https://godbolt.org), a web-based tool that allows users to write, compile, and visualize the assembly output of code in various programming languages; the library is readily available and accessible with the directive `#include <nlohmann/json.hpp>`.
- [**GitHub CodeQL**](https://github.com/github/codeql), a code analysis tool used for identifying security vulnerabilities and bugs in software through semantic queries
+- [**Flutter**](https://github.com/flutter/flutter/blob/master/engine/src/flutter/impeller/compiler/reflector.cc), a UI toolkit for building natively compiled applications for mobile, web, and desktop from a single codebase
+- [**Fraunhofer VVenC**](https://github.com/fraunhoferhhi/vvenc), a fast and efficient encoder for the Versatile Video Coding (H.266/VVC) standard
+- [**GitHub CodeQL**](https://github.com/github/codeql/blob/main/shared/cpp/Diagnostics.h), a code analysis tool used for identifying security vulnerabilities and bugs in software through semantic queries
+- [**GoPro ngfx**](https://github.com/gopro/ngfx), a low-level graphics abstraction and profiling framework developed by GoPro
+- [**gRPC**](https://github.com/grpc/grpc/blob/master/tools/artifact_gen/utils.h), a high-performance universal remote procedure call framework
 - [**Hex-Rays**](https://docs.hex-rays.com/user-guide/user-interface/licenses), a reverse engineering toolset for analyzing and decompiling binaries, primarily used for security research and vulnerability analysis
 - [**ImHex**](https://github.com/WerWolv/ImHex), a hex editor designed for reverse engineering, providing advanced features for data analysis and manipulation
+- [**Intel GITS**](https://github.com/intel/gits), a tool for capturing and replaying graphics API calls for debugging and performance analysis
 - [**Intel GPA Framework**](https://intel.github.io/gpasdk-doc/src/licenses.html), a suite of cross-platform tools for capturing, analyzing, and optimizing graphics applications across different APIs
 - [**Intopix**](https://www.intopix.com/software-licensing), a provider of advanced image processing and compression solutions used in software development and AV workflows
 - [**Java SE**](https://www.oracle.com/a/tech/docs/jdk8-lium.pdf), the core Java platform that provides the libraries and runtime needed to build and run general-purpose Java applications
- [**MKVToolNix**](https://mkvtoolnix.download/doc/README.md), a set of tools for creating, editing, and inspecting MKV (Matroska) multimedia container files
 - [**Meta Yoga**](https://github.com/facebook/yoga), a layout engine that facilitates flexible and efficient user interface design across multiple platforms
- [**NVIDIA Nsight Compute**](https://docs.nvidia.com/nsight-compute/2022.2/pdf/CopyrightAndLicenses.pdf), a performance analysis tool for CUDA applications that provides detailed insights into GPU performance metrics
+- [**MKVToolNix**](https://mkvtoolnix.download/doc/README.md), a set of tools for creating, editing, and inspecting MKV (Matroska) multimedia container files
+- [**MRTech IFF SDK**](https://mr-technologies.com/pub/iff-sdk-manual-2-0-1/iff-sdk-manual-2-0-1.pdf), an image processing SDK for machine vision applications with GPU-accelerated pipelines
+- [**Nix**](https://github.com/NixOS/nix/blob/master/src/nix/build.cc), a purely functional package manager
 - [**Notepad++**](https://github.com/notepad-plus-plus/notepad-plus-plus), a free source code editor that supports various programming languages
+- [**NVIDIA Nsight Compute**](https://docs.nvidia.com/nsight-compute/2022.2/pdf/CopyrightAndLicenses.pdf), a performance analysis tool for CUDA applications that provides detailed insights into GPU performance metrics
+- [**openFrameworks**](https://github.com/openframeworks/openFrameworks/blob/master/libs/openFrameworks/utils/ofJson.h), a community-developed C++ toolkit for creative coding
 - [**OpenRGB**](https://gitlab.com/CalcProgrammer1/OpenRGB), an open source RGB lighting control that doesn't depend on manufacturer software
 - [**OpenTelemetry C++**](https://github.com/open-telemetry/opentelemetry-cpp), a library for collecting and exporting observability data in C++, enabling developers to implement distributed tracing and metrics in their application
+- [**Oracle GraalVM**](https://docs.oracle.com/en/graalvm/jdk/21/docs/licensing-information/), a high-performance JDK distribution with ahead-of-time compilation and polyglot runtime support
+- [**Philips amp-cucumber-cpp-runner**](https://github.com/philips-software/amp-cucumber-cpp-runner), a behaviour-driven development test runner for embedded C++ software developed at Philips
 - [**Qt Creator**](https://doc.qt.io/qtcreator/qtcreator-attribution-json-nlohmann.html), an IDE for developing applications using the Qt application framework
+- [**Qt for MCUs**](https://doc.qt.io/QtForMCUs/quickultralite-attribution-nlohmann-json.html), a graphics framework for building fluid user interfaces on microcontrollers
+- [**React Native**](https://github.com/react/react-native/blob/main/packages/react-native/ReactCxxPlatform/react/devsupport/PackagerConnection.cpp), a framework for building native mobile applications using React
 - [**Scanbot SDK**](https://docs.scanbot.io/barcode-scanner-sdk/web/third-party-libraries/), a software development kit (SDK) that provides tools for integrating advanced document scanning and barcode scanning capabilities into applications
+- [**STMicroelectronics TouchGFX**](https://www.st.com/resource/en/additional_license_terms/additional-license-terms-x-cube-touchgfx.html), a graphical user interface framework shipped with STM32 microcontrollers for building embedded HMIs
+- [**swagger-codegen**](https://github.com/swagger-api/swagger-codegen/blob/master/samples/server/petstore/pistache-server/model/Pet.h), a template-driven engine that generates API clients and server stubs from an OpenAPI specification
+- [**Swoole**](https://github.com/swoole/swoole-src/blob/master/ext-src/swoole_admin_server.cc), a coroutine-based concurrency engine for PHP
+- [**Tracy Profiler**](https://github.com/wolfpld/tracy/blob/master/profiler/src/profiler/TracyLlm.hpp), a real-time frame profiler for games and other applications
+- [**WasmEdge**](https://github.com/WasmEdge/WasmEdge/blob/master/plugins/wasi_nn/GGML/tts/tts_core.cpp), a lightweight WebAssembly runtime for edge and cloud workloads
+- [**x64dbg**](https://github.com/x64dbg/x64dbg/blob/development/src/cross/remote_table/TableRpcData.h), an open source user mode debugger for Windows, aimed at reverse engineering and malware analysis

 ## Machine Learning and AI

+- [**Alibaba MNN**](https://github.com/alibaba/MNN), a lightweight deep learning inference engine for mobile and embedded devices
+- [**AMD Gaia**](https://github.com/amd/gaia), an open-source framework for running generative AI applications locally on AMD hardware
+- [**AMD Vitis AI (VAIP)**](https://github.com/amd/vaip), the execution provider stack that runs AI models on AMD Ryzen AI and adaptive computing devices
 - [**Apple Core ML Tools**](https://github.com/apple/coremltools), a set of tools for converting and configuring machine learning models for deployment in Apple's Core ML framework
 - [**Avular Mobile Robotics**](https://www.avular.com/licenses/nlohmann-json-3.9.1.txt), a platform for developing and deploying mobile robotics solutions
+- [**FunASR**](https://github.com/modelscope/FunASR/blob/main/runtime/http/bin/asr_sessions.h), a speech recognition toolkit for training and deploying end-to-end models
 - [**Google gemma.cpp**](https://github.com/google/gemma.cpp), a lightweight C++ inference engine designed for running AI models from the Gemma family
+- [**Google Magenta The Infinite Crate**](https://github.com/magenta/the-infinite-crate), an open-source generative AI plugin for digital audio workstations from Google's Magenta research team
+- [**GPT4All**](https://github.com/nomic-ai/gpt4all/blob/main/gpt4all-chat/src/tool.h), a desktop application for running local large language models on consumer hardware
+- [**Huawei MindSpore**](https://github.com/mindspore-ai/mindspore/blob/master/Third_Party_Open_Source_Software_Notice), a deep learning framework for training and inference across device, edge, and cloud
+- [**KTransformers**](https://github.com/kvcache-ai/ktransformers/blob/main/archive/csrc/balance_serve/sched/model_config.h), a framework for heterogeneous large language model inference
 - [**llama.cpp**](https://github.com/ggerganov/llama.cpp), a C++ library designed for efficient inference of large language models (LLMs), enabling streamlined integration into applications
+- [**LocalAI**](https://github.com/mudler/LocalAI/blob/master/backend/cpp/ds4/dsml_renderer.cpp), a self-hosted inference engine that exposes local models through an OpenAI-compatible API
 - [**MLX**](https://github.com/ml-explore/mlx), an array framework for machine learning on Apple Silicon
 - [**Mozilla llamafile**](https://github.com/Mozilla-Ocho/llamafile), a tool designed for distributing and executing large language models (LLMs) efficiently using a single file format
 - [**NVIDIA ACE**](https://docs.nvidia.com/ace/latest/index.html), a suite of real-time AI solutions designed for the development of interactive avatars and digital human applications, enabling scalable and sophisticated user interactions
+- [**NVIDIA Instant NGP**](https://github.com/NVlabs/instant-ngp/blob/master/src/nerf_loader.cu), an implementation of instant neural graphics primitives for rapid scene reconstruction
+- [**NVIDIA TensorRT**](https://github.com/NVIDIA/TensorRT), an SDK for high-performance deep learning inference, including its TensorRT-LLM extension for large language models
+- [**NVIDIA TensorRT-LLM**](https://github.com/NVIDIA/TensorRT-LLM/blob/main/cpp/tensorrt_llm/common/safetensors.cpp), a toolkit for optimizing and serving large language model inference on GPUs
+- [**ONNX Runtime**](https://github.com/microsoft/onnxruntime), a cross-platform inference and training accelerator for machine learning models
+- [**OpenVINO**](https://github.com/openvinotoolkit/openvino), Intel's toolkit for optimizing and deploying deep learning inference across CPUs, GPUs, and NPUs
+- [**PaddleOCR**](https://github.com/PaddlePaddle/PaddleOCR/blob/main/deploy/cpp_infer/src/modules/text_detection/result.cc), an optical character recognition toolkit that turns documents and images into structured data
+- [**PaddlePaddle**](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/ap/src/axpr/anf_expr.cc), a deep learning framework for distributed training and inference
 - [**Peer**](https://support.peer.inc/hc/en-us/articles/17261335054235-Licenses), a platform offering personalized AI assistants for interactive learning and creative collaboration
+- [**PyTorch**](https://github.com/pytorch/pytorch), a machine learning framework for building and training neural networks, widely used in research and production
+- [**Qualcomm AI Engine Direct**](https://github.com/qualcomm/qai-appbuilder), a toolchain for building and running generative AI applications on Snapdragon devices
+- [**sherpa-onnx**](https://github.com/k2-fsa/sherpa-onnx/blob/master/sherpa-onnx/csrc/sentence-piece-tokenizer.cc), a speech toolkit for on-device recognition, synthesis and speaker diarization
 - [**stable-diffusion.cpp**](https://github.com/leejet/stable-diffusion.cpp), a C++ implementation of the Stable Diffusion image generation model
 - [**TanvasTouch**](https://tanvas.co/tanvastouch-sdk-third-party-acknowledgments), a software development kit (SDK) that enables developers to create tactile experiences on touchscreens, allowing users to feel textures and physical sensations in a digital environment
 - [**TensorFlow**](https://github.com/tensorflow/tensorflow), a machine learning framework that facilitates the development and training of models, supporting data serialization and efficient data exchange between components
+- [**whisper.cpp**](https://github.com/ggml-org/whisper.cpp), a C++ implementation of OpenAI's Whisper automatic speech recognition model

 ## Scientific Research and Analysis

 - [**BLACK**](https://www.black-sat.org/en/stable/installation/linux.html), a bounded linear temporal logic (LTL) satisfiability checker
+- [**CERN ALICE O2**](https://github.com/AliceO2Group/AliceO2), the online-offline computing framework of the ALICE heavy-ion experiment at the Large Hadron Collider
 - [**CERN Atlas Athena**](https://gitlab.cern.ch/atlas/athena/-/blob/main/Control/PerformanceMonitoring/PerfMonComps/src/PerfMonMTSvc.h), a software framework used in the ATLAS experiment at the Large Hadron Collider (LHC) for performance monitoring
+- [**CERN CMSSW**](https://github.com/cms-sw/cmssw), the offline software framework of the CMS experiment at the Large Hadron Collider
+- [**CERN Gaudi**](https://gitlab.cern.ch/gaudi/Gaudi), the event-processing framework used by the LHCb and ATLAS experiments at the Large Hadron Collider
 - [**ICU**](https://github.com/unicode-org/icu), the International Components for Unicode, a mature library for software globalization and multilingual support
 - [**KAMERA**](https://github.com/Kitware/kamera), a platform for synchronized data collection and real-time deep learning to map marine species like polar bears and seals, aiding Arctic ecosystem research
 - [**KiCad**](https://gitlab.com/kicad/code/kicad/-/tree/master/thirdparty/nlohmann_json), a free and open-source software suite for electronic design automation
+- [**LLNL ROSE**](https://github.com/llnl/rose), a compiler infrastructure from Lawrence Livermore National Laboratory for building source-to-source program analysis and transformation tools
 - [**Maple**](https://www.maplesoft.com/support/help/Maple/view.aspx?path=copyright), a symbolic and numeric computing environment for advanced mathematical modeling and analysis
 - [**MeVisLab**](https://mevislabdownloads.mevis.de/docs/current/MeVis/ThirdParty/Documentation/Publish/ThirdPartyReference/index.html), a software framework for medical image processing and visualization.
+- [**MITK**](https://github.com/MITK/MITK), the Medical Imaging Interaction Toolkit, a framework for developing interactive medical image processing software
 - [**OpenPMD API**](https://openpmd-api.readthedocs.io/en/0.8.0-alpha/backends/json.html), a versatile programming interface for accessing and managing scientific data, designed to facilitate the efficient storage, retrieval, and sharing of simulation data across various applications and platforms
+- [**ORNL DataFed**](https://github.com/ORNL/DataFed), a federated scientific data management system developed at Oak Ridge National Laboratory
 - [**ParaView**](https://github.com/Kitware/ParaView), an open-source tool for large-scale data visualization and analysis across various scientific domains
 - [**QGIS**](https://gitlab.b-data.ch/qgis/qgis/-/blob/backport-57658-to-release-3_34/external/nlohmann/json.hpp), a free and open-source geographic information system (GIS) application that allows users to create, edit, visualize, and analyze geospatial data across a variety of formats
- [**VTK**](https://github.com/Kitware/VTK), a software library for 3D computer graphics, image processing, and visualization
+- [**Sandia InterSpec**](https://github.com/sandialabs/InterSpec), spectral radiation analysis software from Sandia National Laboratories for identifying radioactive isotopes
 - [**VolView**](https://github.com/Kitware/VolView), a lightweight application for interactive visualization and analysis of 3D medical imaging data.
+- [**VTK**](https://github.com/Kitware/VTK), a software library for 3D computer graphics, image processing, and visualization

 ## Business and Productivity Software

 - [**ArcGIS PRO**](https://www.esri.com/content/dam/esrisites/en-us/media/legal/open-source-acknowledgements/arcgis-pro-2-8-attribution-report.html), a desktop geographic information system (GIS) application developed by Esri for mapping and spatial analysis
 - [**Autodesk Desktop**](https://damassets.autodesk.net/content/dam/autodesk/www/Company/legal-notices-trademarks/autodesk-desktop-platform-components/internal-autodesk-components-web-page-2023.pdf), a software platform developed by Autodesk for creating and managing desktop applications and services
 - [**Check Point**](https://www.checkpoint.com/about-us/copyright-and-trademarks/), a cybersecurity company specializing in threat prevention and network security solutions, offering a range of products designed to protect enterprises from cyber threats and ensure data integrity
+- [**EasyEffects**](https://github.com/wwmm/easyeffects/blob/master/src/presets_manager.hpp), an audio effects processor for PipeWire offering limiting, compression and equalization
+- [**espanso**](https://github.com/espanso/espanso/blob/dev/espanso-ui/src/win32/native.cpp), a cross-platform text expander
+- [**Karabiner-Elements**](https://github.com/pqrs-org/Karabiner-Elements/blob/main/src/share/app_icon.hpp), a keyboard customizer for macOS
+- [**MacType**](https://github.com/snowie2000/mactype/blob/directwrite/settings.h), a font rendering engine for Windows
+- [**magicplan**](https://help.magicplan.app/acknowledgments), a mobile application for creating floor plans and interior designs using augmented reality
 - [**Microsoft Office for Mac**](https://officecdnmac.microsoft.com/pr/legal/mac/OfficeforMacAttributions.html), a suite of productivity applications developed by Microsoft for macOS, including tools for word processing, spreadsheets, and presentations
 - [**Microsoft Teams**](https://www.microsoft.com/microsoft-teams/), a team collaboration application offering workspace chat and video conferencing, file storage, and integration of proprietary and third-party applications and services
+- [**MuseScore**](https://github.com/musescore/MuseScore), a free and open-source music notation and composition application
+- [**NanaZip**](https://github.com/M2Team/NanaZip/blob/main/NanaZip.Codecs/NanaZip.Codecs.Archive.ElectronAsar.cpp), a 7-Zip derivative built for modern Windows
 - [**Nexthink Infinity**](https://docs.nexthink.com/legal/services-terms/experience-open-source-software-licenses/infinity-2022.8-software-licenses), a digital employee experience management platform for monitoring and improving IT performance
 - [**Sophos Connect Client**](https://docs.sophos.com/nsg/licenses/SophosConnect/SophosConnectAttribution.html), a secure VPN client from Sophos that allows remote users to connect to their corporate network, ensuring secure access to resources and data
 - [**Stonebranch**](https://stonebranchdocs.atlassian.net/wiki/spaces/UA77/pages/799545647/Licenses+for+Third-Party+Libraries), a cloud-based cybersecurity solution that integrates backup, disaster recovery, and cybersecurity features to protect data and ensure business continuity for organizations
 - [**Tablecruncher**](https://tablecruncher.com/), a data analysis tool that allows users to import, analyze, and visualize spreadsheet data, offering interactive features for better insights and decision-making
- [**magicplan**](https://help.magicplan.app/acknowledgments), a mobile application for creating floor plans and interior designs using augmented reality
+- [**VNote**](https://github.com/vnotex/vnote/blob/master/src/core/services/notebookcoreservice.cpp), a Markdown-based note-taking application written in C++

 ## Databases and Big Data

 - [**ADIOS2**](https://code.ornl.gov/ecpcitest/adios2/-/tree/pr4285_FFSUpstream/thirdparty/nlohmann_json?ref_type=heads), a data management framework designed for high-performance input and output operations
+- [**Apache Doris**](https://github.com/apache/doris/blob/master/be/src/runtime/be_proc_monitor.cpp), a real-time analytical database for high-concurrency queries
+- [**Claris FileMaker Server**](https://www.claris.com/company/legal/docs/acknowledgements/filemaker-server-macwin/claris_fms2025_acknowledgements_en.pdf), the server platform hosting FileMaker custom apps and databases, developed by Apple subsidiary Claris
+- [**ClickHouse**](https://github.com/ClickHouse/ClickHouse), a column-oriented database management system for real-time analytical queries
 - [**Cribl Stream**](https://docs.cribl.io/stream/third-party-current-list/), a real-time data processing platform that enables organizations to collect, route, and transform observability data, enhancing visibility and insights into their systems
 - [**DB Browser for SQLite**](https://github.com/sqlitebrowser/sqlitebrowser), a visual open-source tool for creating, designing, and editing SQLite database files
+- [**Manticore Search**](https://github.com/manticoresoftware/manticoresearch/blob/main/src/searchdhttpcompat.cpp), a database for search, offering full-text and vector queries
+- [**Milvus**](https://github.com/milvus-io/milvus/blob/master/internal/core/src/query/PlanImpl.h), a cloud-native vector database built for embedding similarity search
+- [**MongoDB**](https://github.com/mongodb/mongo/blob/master/src/mongo/replay/config_handler.cpp), a general-purpose document database
 - [**MySQL Connector/C++**](https://docs.oracle.com/cd/E17952_01/connector-cpp-9.1-license-com-en/license-opentelemetry-cpp-com.html), a C++ library for connecting and interacting with MySQL databases
 - [**MySQL NDB Cluster**](https://downloads.mysql.com/docs/licenses/cluster-9.0-com-en.pdf), a distributed database system that provides high availability and scalability for MySQL databases
 - [**MySQL Shell**](https://downloads.mysql.com/docs/licenses/mysql-shell-8.0-gpl-en.pdf), an advanced client and code editor for interacting with MySQL servers, supporting SQL, Python, and JavaScript
- [**PrestoDB**](https://github.com/prestodb/presto), a distributed SQL query engine designed for large-scale data analytics, originally developed by Facebook
+- [**PrestoDB**](https://github.com/prestodb/presto/blob/master/presto-native-execution/presto_cpp/main/Announcer.cpp), a distributed SQL query engine designed for large-scale data analytics, originally developed by Facebook
 - [**ROOT Data Analysis Framework**](https://root.cern/doc/v614/classnlohmann_1_1basic__json.html), an open-source data analysis framework widely used in high-energy physics and other fields for data processing and visualization
+- [**Typesense**](https://github.com/typesense/typesense/blob/v31/include/join.h), an open source typo-tolerant search engine
+- [**Vearch**](https://github.com/jd-opensource/vearch), a distributed vector database developed at JD.com for similarity search and retrieval-augmented generation
 - [**WiredTiger**](https://github.com/wiredtiger/wiredtiger), a high-performance storage engine for databases, offering support for compression, concurrency, and checkpointing

 ## Simulation and Modeling

+- [**Adobe Lagrange**](https://github.com/adobe/lagrange), a geometry processing library developed by Adobe for mesh manipulation and analysis
 - [**Arcturus HoloSuite**](https://www.datocms-assets.com/104353/1698904597-holosuite-third-party-software-credits-and-attributions-2.pdf), a software toolset for capturing, editing, and streaming volumetric video, featuring advanced compression technologies for high-quality 3D content creation
 - [**azul**](https://pure.tudelft.nl/ws/files/85338589/tgis.12673.pdf), a fast and efficient 3D city model viewer designed for visualizing urban environments and spatial data
+- [**Bambu Studio**](https://github.com/bambulab/BambuStudio), a slicing and print management application for Bambu Lab 3D printers
 - [**Blender**](https://projects.blender.org/blender/blender/search?q=nlohmann), a free and open-source 3D creation suite for modeling, animation, rendering, and more
 - [**cpplot**](https://cpplot.readthedocs.io/en/latest/library_api/function_eigen_8h_1ac080eac0541014c5892a55e41bf785e6.html), a library for creating interactive graphs and charts in C++, which can be viewed in web browsers
 - [**Foundry Nuke**](https://learn.foundry.com/nuke/content/misc/studio_third_party_libraries.html), a powerful node-based digital compositing and visual effects application used in film and television post-production
+- [**FreeCAD**](https://github.com/FreeCAD/FreeCAD), a free and open-source parametric 3D CAD modeler for product design and engineering
 - [**GAMS**](https://www.gams.com/47/docs/THIRDPARTY.html), a high-performance mathematical modeling system for optimization and decision support
+- [**Keysight WirelessPro**](https://docs.keysight.com/display/engdocwirelesspro/WirelessPro+2026+Release+Notes), a simulation platform for 5G, 5G-Advanced, and 6G cellular network research
 - [**Kitware SMTK**](https://github.com/Kitware/SMTK), a software toolkit for managing simulation models and workflows in scientific and engineering applications
 - [**M-Star**](https://docs.mstarcfd.com/3_Licensing/thirdparty-licenses.html), a computational fluid dynamics software for simulating and analyzing fluid flow
 - [**MapleSim CAD Toolbox**](https://www.maplesoft.com/support/help/MapleSim/view.aspx?path=CADToolbox/copyright), a software extension for MapleSim that integrates CAD models, allowing users to import, manipulate, and analyze 3D CAD data within the MapleSim environment for enhanced modeling and simulation
+- [**Microsoft AirSim**](https://github.com/microsoft/AirSim/blob/main/AirLib/include/common/Settings.hpp), a simulator for autonomous vehicles and drones built on Unreal Engine
 - [**NVIDIA Omniverse**](https://docs.omniverse.nvidia.com/composer/latest/common/product-licenses/usd-explorer/usd-explorer-2023.2.0-licenses-manifest.html), a platform for 3D content creation and collaboration that enables real-time simulations and interactive experiences across various industries
+- [**OpenSCAD**](https://github.com/openscad/openscad/blob/master/src/core/AIClient.cc), a script-driven solid 3D CAD modeller
+- [**OrcaSlicer**](https://github.com/SoftFever/OrcaSlicer), an open-source slicer supporting a wide range of consumer 3D printers
 - [**Pixar Renderman**](https://rmanwiki-26.pixar.com/space/REN26/19662083/Legal+Notice), a photorealistic 3D rendering software developed by Pixar, widely used in the film industry for creating high-quality visual effects and animations
+- [**PrusaSlicer**](https://github.com/prusa3d/PrusaSlicer), the slicing software developed by Prusa Research for its 3D printers
 - [**ROS - Robot Operating System**](http://docs.ros.org/en/noetic/api/behaviortree_cpp/html/json_8hpp_source.html), a set of software libraries and tools that assist in developing robot applications
 - [**UBS**](https://www.ubs.com/), a multinational financial services and banking company

@@ -161,18 +278,31 @@ the result of an internet search. If you know further customers of the library,
 - [**Acronis Cyber Protect Cloud**](https://care.acronis.com/s/article/59533-Third-party-software-used-in-Acronis-Cyber-Protect-Cloud?language=en_US), an all-in-one data protection solution that combines backup, disaster recovery, and cybersecurity to safeguard business data from threats like ransomware
 - [**Baereos**](https://gitlab.tiger-computing.co.uk/packages/bareos/-/blob/tiger/bullseye/third-party/CLI11/examples/json.cpp), a backup solution that provides data protection and recovery options for various environments, including physical and virtual systems
 - [**Bitdefender Home Scanner**](https://www.bitdefender.de/site/Main/view/home-scanner-open-source.html), a tool from Bitdefender that scans devices for malware and security threats, providing a safeguard against potential online dangers
+- [**Cisco MLS++**](https://github.com/cisco/mlspp), an implementation of the Messaging Layer Security protocol for end-to-end encrypted group messaging
 - [**Citrix Provisioning**](https://docs.citrix.com/en-us/provisioning/2203-ltsr/downloads/pvs-third-party-notices-2203.pdf), a solution that streamlines the delivery of virtual desktops and applications by allowing administrators to manage and provision resources efficiently across multiple environments
 - [**Citrix Virtual Apps and Desktops**](https://docs.citrix.com/en-us/citrix-virtual-apps-desktops/2305/downloads/third-party-notices-apps-and-desktops.pdf), a solution from Citrix that delivers virtual apps and desktops
 - [**Cyberarc**](https://docs.cyberark.com/Downloads/Legal/Privileged%20Session%20Manager%20for%20SSH%20Third-Party%20Notices.pdf), a security solution that specializes in privileged access management, enabling organizations to control and monitor access to critical systems and data, thereby enhancing overall cybersecurity posture
+- [**Deutsche Telekom sysrepo-plugins**](https://github.com/telekom/sysrepo-plugins), a collection of YANG datastore plugins used to manage network devices
 - [**Egnyte Desktop**](https://helpdesk.egnyte.com/hc/en-us/articles/360007071732-Third-Party-Software-Acknowledgements), a secure cloud storage solution designed for businesses, enabling file sharing, collaboration, and data management across teams while ensuring compliance and data protection
 - [**Elster**](https://www.secunet.com/en/about-us/press/article/elstersecure-bietet-komfortablen-login-ohne-passwort-dank-secunet-protect4use), a digital platform developed by German tax authorities for secure and efficient electronic tax filing and management using secunet protect4use
+- [**Envoy**](https://github.com/envoyproxy/envoy), a cloud-native edge and service proxy that forms the data plane of many service meshes
 - [**Ethereum Solidity**](https://github.com/ethereum/solidity), a high-level, object-oriented programming language designed for implementing smart contracts on the Ethereum platform
+- [**gVisor**](https://github.com/google/gvisor), an application kernel that provides a secure sandbox for running untrusted containers
+- [**IBM Storage Virtualize**](https://public.dhe.ibm.com/systems/support/warranty/pdfs/stgoilc/SV_for_FS_7300_v8_7_0_Base_OILC.pdf), the software powering IBM FlashSystem enterprise storage arrays
 - [**Inciga**](https://fossies.org/linux/icinga2/third-party/nlohmann_json/json.hpp), a monitoring tool for IT infrastructure, designed to provide insights into system performance and availability through customizable dashboards and alerts
 - [**Intel Accelerator Management Daemon for VMware ESXi**](https://downloadmirror.intel.com/772507/THIRD-PARTY.txt), a management tool designed for monitoring and controlling Intel hardware accelerators within VMware ESXi environments, optimizing performance and resource allocation
 - [**Juniper Identity Management Service**](https://www.juniper.net/documentation/us/en/software/jims/jims-guide/jims-guide.pdf)
+- [**Meta FBOSS**](https://github.com/facebook/fboss), the software stack that controls the network switches in Meta's data centers
 - [**Microsoft Azure IoT SDK**](https://library.e.abb.com/public/2779c5f85f30484192eb3cb3f666a201/IP%20Gateway%20Open%20License%20Declaration_9AKK108467A4095_Rev_C.pdf), a collection of tools and libraries to help developers connect, build, and deploy Internet of Things (IoT) solutions on the Azure cloud platform
+- [**Microsoft Confidential Consortium Framework**](https://github.com/microsoft/CCF), a framework for building secure, highly available applications on trusted execution environments
 - [**Microsoft WinGet**](https://github.com/microsoft/winget-cli), a command-line utility included in the Windows Package Manager
+- [**Mitsubishi Electric SECS/GEM**](https://dl.mitsubishielectric.com/dl/fa/document/manual/plc/sh082483eng/sh082483engi.pdf), the semiconductor equipment communication software running on Mitsubishi Electric C Controller and C intelligent function modules
+- [**Moxa**](https://www.moxa.com/getmedia/fbe2a0c7-8dda-4b5b-a501-15e45adebb1f/moxa-foss-statement-for-da-720-series-win-10-ltsc-21h2-declaration-v1.0.pdf), a provider of industrial networking, computing, and automation infrastructure
 - [**plexusAV**](https://www.sisme.com/media/10994/manual_plexusav-p-avn-4-form8244-c.pdf), a high-performance AV-over-IP transceiver device capable of video encoding and decoding using the IPMX standard
 - [**Pointr**](https://docs-dev.pointr.tech/docs/8.x/Developer%20Portal/Open%20Source%20Licenses/), a platform for indoor positioning and navigation solutions, offering tools and SDKs for developers to create location-based applications
 - [**secunet protect4use**](https://www.secunet.com/en/about-us/press/article/elstersecure-bietet-komfortablen-login-ohne-passwort-dank-secunet-protect4use), a secure, passwordless multifactor authentication solution that transforms smartphones into digital keyrings, ensuring high security for online services and digital identities
 - [**Sencore MRD 7000**](https://www.foccusdigital.com/wp-content/uploads/2025/03/MRD-7000-Manual-8175V.pdf), a professional multi-channel receiver and decoder supporting UHD and HD stream decoding
+- [**Siemens SINEC**](https://cache.industry.siemens.com/dl/files/917/109974917/att_1298783/v2/OSS_SINEC-NMS_99.pdf), a family of network management and infrastructure services for industrial networks
+- [**Toshiba Industrial Servers**](https://www.global.toshiba/content/dam/toshiba/jp/products-solutions/industrial/computer/product/server/fs20000r/pdf/FS20000R_OSS_License_6E8C5817_rev0.pdf), the FS20000R series of industrial servers for factory automation and control systems
+- [**Wazuh**](https://github.com/wazuh/wazuh/blob/main/src/data_provider/src/sysInfo.cpp), a security platform for threat detection, integrity monitoring and incident response
+- [**ZeroTier**](https://github.com/zerotier/ZeroTierOne/blob/dev/osdep/OSUtils.hpp), a software-defined networking service that creates virtual Ethernet networks
@@ -868,6 +868,12 @@ The size of an array or object in a [binary format](../features/binary_formats/i
 the size following `#` for [UBJSON](../features/binary_formats/ubjson.md)/[BJData](../features/binary_formats/bjdata.md),
 or the encoded length for [CBOR](../features/binary_formats/cbor.md).

+The exception is also thrown for a [UBJSON](../features/binary_formats/ubjson.md) array of a type that is encoded by its
+marker alone (`Z`, `T` or `F`) whose declared count exceeds 1,048,576. Such an array has no payload, so its count alone
+decides how much memory is allocated, and a handful of bytes would otherwise describe billions of values.
+[`to_ubjson`](../api/basic_json/to_ubjson.md) writes longer arrays of these types without the size and type annotation,
+so any value it produces can still be read back.
+
 !!! failure "Example messages"

    ```
@@ -879,6 +885,9 @@ or the encoded length for [CBOR](../features/binary_formats/cbor.md).
    ```
    [json.exception.out_of_range.408] syntax error while parsing CBOR size: excessive map size
    ```
+    ```
+    [json.exception.out_of_range.408] syntax error while parsing UBJSON size: excessive array size
+    ```

 ### json.exception.out_of_range.409

@@ -58,6 +58,26 @@ inline bool little_endianness(int num = 1) noexcept
    return *reinterpret_cast<char*>(&num) == 1;
 }

+/*!
+@brief largest element count accepted for a UBJSON container of a valueless type
+
+An element of type 'Z' (null), 'T' (true) or 'F' (false) is encoded by its
+type marker alone, so an optimized container of one of those types has no
+payload at all and its declared count is the only thing that decides how much
+is allocated: `[$Z#L` followed by a large count turns some ten bytes of input
+into that many values (see #2793, which reports 35 GB and 150 seconds). Every
+other type costs at least one byte per element and is bounded by the end of
+the input.
+
+This is a sanity bound rather than a security boundary, and it is far above
+any container met in practice. @ref binary_writer falls back to the
+unoptimized encoding for longer containers, so that a value serialized by
+this library can always be read back.
+
+@sa https://github.com/nlohmann/json/issues/2793
+*/
+JSON_INLINE_VARIABLE constexpr std::size_t max_valueless_container_size = 1 << 20;
+
 ///////////////////
 // binary reader //
 ///////////////////
@@ -996,23 +1016,21 @@ class binary_reader
    }

    /*!
-    @brief reads a CBOR string
+    @brief reads a definite-length CBOR string

-    This function first reads starting bytes to determine the expected
-    string length and then copies this number of bytes into a string.
-    Additionally, CBOR's strings with indefinite lengths are supported.
+    Reads everything @ref get_cbor_string accepts except the indefinite-length
+    form, which that function handles itself. The bytes are appended to @a
+    result, so consecutive chunks of an indefinite-length string can be read
+    into the same string.

-    @param[out] result  created string
+    @param[out] result  string the bytes are appended to

    @return whether string creation completed
-    */
-    bool get_cbor_string(string_t& result)
-    {
-        if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "string")))
-        {
-            return false;
-        }

+    @pre @a current is not EOF
+    */
+    bool get_cbor_string_chunk(string_t& result)
+    {
        switch (current)
        {
            // UTF-8 string (0x00..0x17 bytes follow)
@@ -1068,20 +1086,6 @@ class binary_reader
                return get_number(input_format_t::cbor, len) && get_string(input_format_t::cbor, len, result);
            }

-            case 0x7F: // UTF-8 string (indefinite length)
-            {
-                while (get() != 0xFF)
-                {
-                    string_t chunk;
-                    if (!get_cbor_string(chunk))
-                    {
-                        return false;
-                    }
-                    result.append(chunk);
-                }
-                return true;
-            }
-
            default:
            {
                auto last_token = get_token_string();
@@ -1092,23 +1096,82 @@ class binary_reader
    }

    /*!
-    @brief reads a CBOR byte array
+    @brief reads a CBOR string

    This function first reads starting bytes to determine the expected
-    byte array length and then copies this number of bytes into the byte array.
-    Additionally, CBOR's byte arrays with indefinite lengths are supported.
+    string length and then copies this number of bytes into a string.
+    Additionally, CBOR's strings with indefinite lengths are supported.

-    @param[out] result  created byte array
+    @param[out] result  created string
+
+    @return whether string creation completed
+    */
+    bool get_cbor_string(string_t& result)
+    {
+        // number of indefinite-length strings that have been opened and not
+        // closed yet. RFC 8949, Section 3.2.3 does not permit nesting them,
+        // but this reader has always accepted it, so the open levels are
+        // counted instead of recursed through, which overflowed the stack for
+        // an input of repeated 0x7F bytes (see #5104). Every chunk is appended
+        // to the same result, so no per-level state is needed.
+        std::size_t open = 0;
+
+        while (true)
+        {
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "string")))
+            {
+                return false;
+            }
+
+            if (current == 0x7F) // UTF-8 string (indefinite length)
+            {
+                ++open;
+                get();
+                continue;
+            }
+
+            // a break marker closes the innermost indefinite-length string;
+            // outside of one it is not a string and falls through to the error
+            if (open != 0 && current == 0xFF)
+            {
+                if (--open == 0)
+                {
+                    return true;
+                }
+                get();
+                continue;
+            }
+
+            if (JSON_HEDLEY_UNLIKELY(!get_cbor_string_chunk(result)))
+            {
+                return false;
+            }
+
+            if (open == 0)
+            {
+                return true;
+            }
+
+            get();
+        }
+    }
+
+    /*!
+    @brief reads a definite-length CBOR byte array
+
+    Reads everything @ref get_cbor_binary accepts except the indefinite-length
+    form, which that function handles itself. The bytes are appended to @a
+    result, so consecutive chunks of an indefinite-length byte array can be
+    read into the same byte array.
+
+    @param[out] result  byte array the bytes are appended to

    @return whether byte array creation completed
-    */
-    bool get_cbor_binary(binary_t& result)
-    {
-        if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "binary")))
-        {
-            return false;
-        }

+    @pre @a current is not EOF
+    */
+    bool get_cbor_binary_chunk(binary_t& result)
+    {
        switch (current)
        {
            // Binary data (0x00..0x17 bytes follow)
@@ -1168,20 +1231,6 @@ class binary_reader
                       get_binary(input_format_t::cbor, len, result);
            }

-            case 0x5F: // Binary data (indefinite length)
-            {
-                while (get() != 0xFF)
-                {
-                    binary_t chunk;
-                    if (!get_cbor_binary(chunk))
-                    {
-                        return false;
-                    }
-                    result.insert(result.end(), chunk.begin(), chunk.end());
-                }
-                return true;
-            }
-
            default:
            {
                auto last_token = get_token_string();
@@ -1191,6 +1240,63 @@ class binary_reader
        }
    }

+    /*!
+    @brief reads a CBOR byte array
+
+    This function first reads starting bytes to determine the expected
+    byte array length and then copies this number of bytes into the byte array.
+    Additionally, CBOR's byte arrays with indefinite lengths are supported.
+
+    @param[out] result  created byte array
+
+    @return whether byte array creation completed
+    */
+    bool get_cbor_binary(binary_t& result)
+    {
+        // the open indefinite-length byte arrays are counted rather than
+        // recursed through, for the reason given in @ref get_cbor_string
+        std::size_t open = 0;
+
+        while (true)
+        {
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "binary")))
+            {
+                return false;
+            }
+
+            if (current == 0x5F) // Binary data (indefinite length)
+            {
+                ++open;
+                get();
+                continue;
+            }
+
+            // a break marker closes the innermost indefinite-length byte
+            // array; outside of one it falls through to the error below
+            if (open != 0 && current == 0xFF)
+            {
+                if (--open == 0)
+                {
+                    return true;
+                }
+                get();
+                continue;
+            }
+
+            if (JSON_HEDLEY_UNLIKELY(!get_cbor_binary_chunk(result)))
+            {
+                return false;
+            }
+
+            if (open == 0)
+            {
+                return true;
+            }
+
+            get();
+        }
+    }
+
    /*!
    @brief narrow a definite CBOR array/map length to std::size_t

@@ -2391,7 +2497,12 @@ class binary_reader
    {
        result.first = npos; // size
        result.second = 0; // type
-        bool is_ndarray = false;
+        // seed the flag with the caller's context: inside an ndarray dimension
+        // vector another ndarray is not allowed, and get_ubjson_size_value()
+        // rejects it up front instead of reading it and reporting afterwards.
+        // Seeding it with `false` made every '#' of a "[#[#[..." chain descend
+        // another level, which overflowed the stack (see #5104).
+        bool is_ndarray = inside_ndarray;

        get_ignore_noop();

@@ -2424,13 +2535,11 @@ class binary_reader
            }

            const bool is_error = get_ubjson_size_value(result.first, is_ndarray);
-            if (input_format == input_format_t::bjdata && is_ndarray)
+            // an ndarray was read here only if the flag flipped; when it was
+            // seeded true, get_ubjson_size_value() already rejected the nested
+            // dimension vector
+            if (input_format == input_format_t::bjdata && is_ndarray && !inside_ndarray)
            {
-                if (inside_ndarray)
-                {
-                    return sax->parse_error(chars_read, get_token_string(), parse_error::create(112, chars_read,
-                                            exception_message(input_format, "ndarray can not be recursive", "size"), nullptr));
-                }
                result.second |= (1 << 8); // use bit 8 to indicate ndarray, all UBJSON and BJData markers should be ASCII letters
            }
            return is_error;
@@ -2439,7 +2548,7 @@ class binary_reader
        if (current == '#')
        {
            const bool is_error = get_ubjson_size_value(result.first, is_ndarray);
-            if (input_format == input_format_t::bjdata && is_ndarray)
+            if (input_format == input_format_t::bjdata && is_ndarray && !inside_ndarray)
            {
                return sax->parse_error(chars_read, get_token_string(), parse_error::create(112, chars_read,
                                        exception_message(input_format, "ndarray requires both type and size", "size"), nullptr));
@@ -2710,6 +2819,17 @@ class binary_reader

        if (size_and_type.first != npos)
        {
+            // reading an element of a valueless type consumes no input, so the
+            // declared count alone decides how much is allocated; the check is
+            // made before the start event so that no container is opened that
+            // is then abandoned. See @ref max_valueless_container_size.
+            if (JSON_HEDLEY_UNLIKELY((size_and_type.second == 'Z' || size_and_type.second == 'T' || size_and_type.second == 'F')
+                                     && size_and_type.first > max_valueless_container_size))
+            {
+                return sax->parse_error(chars_read, get_token_string(), out_of_range::create(408,
+                                        exception_message(input_format, "excessive array size", "size"), nullptr));
+            }
+
            if (JSON_HEDLEY_UNLIKELY(!sax->start_array(size_and_type.first)))
            {
                return false;
@@ -826,7 +826,17 @@ class binary_writer

                    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()))
+                    // 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
+                    // those, at one byte per element, so the result can be read back.
+                    // Objects are not affected: every element is preceded by its key.
+                    const bool valueless_type = (first_prefix == 'Z' || first_prefix == 'T' || first_prefix == 'F');
+                    const bool excessive_valueless = valueless_type
+                                                     && 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()))
                    {
                        prefix_required = false;
                        oa->write_character(to_char_type('$'));
@@ -1668,6 +1678,15 @@ class binary_writer
        CharType dtype = it->second;

        key = "_ArraySize_";
+        // the dimensions are written verbatim as the header length below, so a
+        // value that is not an array cannot produce a valid one: null emits 'Z'
+        // and an object emits '{', neither of which a reader accepts after '#'.
+        // Such an object is not a valid ndarray and falls back to a plain object.
+        if (!value.at(key).is_array())
+        {
+            return true;
+        }
+
        std::size_t len = (value.at(key).empty() ? 0 : 1);
        for (const auto& el : value.at(key))
        {
@@ -9,9 +9,12 @@
 #pragma once

 #include <array> // array
+#include <cmath> // isnan, ldexp, trunc
 #include <cstddef> // size_t
 #include <cstdint> // uint8_t
+#include <limits> // numeric_limits
 #include <string> // string
+#include <type_traits> // is_signed

 #include <nlohmann/detail/macro_scope.hpp>
 #if JSON_HAS_THREE_WAY_COMPARISON
@@ -114,5 +117,67 @@ inline bool operator<(const value_t lhs, const value_t rhs) noexcept
 }
 #endif

+
+/*!
+@brief compare an integer with a floating point number without precision loss
+
+Widening the integer to the floating point type loses precision beyond the
+float's mantissa, which makes equality intransitive: both 2^63-2 and 2^63-1
+round to 2^63, so each compares equal to that float while differing from each
+other. Ordering built on that is not a strict weak ordering, so sorting such
+values, or using them as keys in an ordered container, is undefined behavior.
+
+Returns a value to be compared against zero with the original operator, which
+reproduces the exact ordering. A NaN operand is returned as is, so comparing it
+against zero keeps NaN's semantics: false for the relational operators and
+unordered for `<=>`.
+*/
+template<typename IntegerType, typename FloatType>
+FloatType compare_integer_with_float(const IntegerType i, const FloatType f) noexcept
+{
+    const auto ordered = [](int c) noexcept
+    {
+        return static_cast<FloatType>(c);
+    };
+
+    if (std::isnan(f))
+    {
+        return f;
+    }
+
+    // values of IntegerType lie in [-bound, bound) when signed and in
+    // [0, bound) when unsigned; digits excludes the sign bit, so bound is a
+    // power of two that the float represents exactly
+    const FloatType bound = std::ldexp(static_cast<FloatType>(1), std::numeric_limits<IntegerType>::digits);
+    if (f >= bound)
+    {
+        return ordered(-1);
+    }
+    if (std::is_signed<IntegerType>::value ? (f < -bound) : (f < static_cast<FloatType>(0)))
+    {
+        return ordered(1);
+    }
+
+    // f is now within the integer's range, so truncating it is exact
+    const FloatType truncated = std::trunc(f);
+    const auto as_integer = static_cast<IntegerType>(truncated);
+    if (i != as_integer)
+    {
+        return ordered(i < as_integer ? -1 : 1);
+    }
+
+    // the integer parts agree, so any fractional part decides
+    const FloatType fraction = f - truncated;
+    if (fraction > static_cast<FloatType>(0))
+    {
+        return ordered(-1);
+    }
+    if (fraction < static_cast<FloatType>(0))
+    {
+        return ordered(1);
+    }
+    return ordered(0);
+}
+
 }  // namespace detail
 NLOHMANN_JSON_NAMESPACE_END
@@ -3703,19 +3703,19 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
    }                                                                                                    \
    else if (lhs_type == value_t::number_integer && rhs_type == value_t::number_float)                   \
    {                                                                                                    \
-        return static_cast<number_float_t>(lhs.m_data.m_value.number_integer) op rhs.m_data.m_value.number_float;      \
+        return (detail::compare_integer_with_float(lhs.m_data.m_value.number_integer, rhs.m_data.m_value.number_float)) op (static_cast<number_float_t>(0)); \
    }                                                                                                    \
    else if (lhs_type == value_t::number_float && rhs_type == value_t::number_integer)                   \
    {                                                                                                    \
-        return lhs.m_data.m_value.number_float op static_cast<number_float_t>(rhs.m_data.m_value.number_integer);      \
+        return (static_cast<number_float_t>(0)) op (detail::compare_integer_with_float(rhs.m_data.m_value.number_integer, lhs.m_data.m_value.number_float)); \
    }                                                                                                    \
    else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_float)                  \
    {                                                                                                    \
-        return static_cast<number_float_t>(lhs.m_data.m_value.number_unsigned) op rhs.m_data.m_value.number_float;     \
+        return (detail::compare_integer_with_float(lhs.m_data.m_value.number_unsigned, rhs.m_data.m_value.number_float)) op (static_cast<number_float_t>(0)); \
    }                                                                                                    \
    else if (lhs_type == value_t::number_float && rhs_type == value_t::number_unsigned)                  \
    {                                                                                                    \
-        return lhs.m_data.m_value.number_float op static_cast<number_float_t>(rhs.m_data.m_value.number_unsigned);     \
+        return (static_cast<number_float_t>(0)) op (detail::compare_integer_with_float(rhs.m_data.m_value.number_unsigned, lhs.m_data.m_value.number_float)); \
    }                                                                                                    \
    else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_integer)                \
    {                                                                                                    \
@@ -4474,7 +4474,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::forward<InputType>(i));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::cbor).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in CBOR format (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -4491,7 +4491,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::move(first), std::move(last));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::cbor).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    template<typename T>
@@ -4517,7 +4517,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::cbor).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in MessagePack format
@@ -4532,7 +4532,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::forward<InputType>(i));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::msgpack).sax_parse(input_format_t::msgpack, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in MessagePack format (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -4548,7 +4548,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::move(first), std::move(last));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::msgpack).sax_parse(input_format_t::msgpack, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    template<typename T>
@@ -4572,7 +4572,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::msgpack).sax_parse(input_format_t::msgpack, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in UBJSON format
@@ -4587,7 +4587,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::forward<InputType>(i));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::ubjson).sax_parse(input_format_t::ubjson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in UBJSON format (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -4603,7 +4603,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::move(first), std::move(last));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::ubjson).sax_parse(input_format_t::ubjson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    template<typename T>
@@ -4627,7 +4627,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::ubjson).sax_parse(input_format_t::ubjson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in BJData format
@@ -4642,7 +4642,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::forward<InputType>(i));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::bjdata).sax_parse(input_format_t::bjdata, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in BJData format (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -4658,7 +4658,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::move(first), std::move(last));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::bjdata).sax_parse(input_format_t::bjdata, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in BSON format
@@ -4673,7 +4673,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::forward<InputType>(i));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::bson).sax_parse(input_format_t::bson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in BSON format (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -4689,7 +4689,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::move(first), std::move(last));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::bson).sax_parse(input_format_t::bson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    template<typename T>
@@ -4713,7 +4713,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::bson).sax_parse(input_format_t::bson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }
    /// @}

@@ -5128,19 +5128,17 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
                // We now reached the end of at least one array
                // in a second pass, traverse the remaining elements

-                // remove my remaining elements
-                const auto end_index = static_cast<difference_type>(result.size());
-                while (i < source.size())
+                // remove my remaining elements, highest index first; appending
+                // in that order avoids the quadratic reinsertion done before
+                for (std::size_t j = source.size(); j > i; --j)
                {
-                    // add operations in reverse order to avoid invalid
-                    // indices
-                    result.insert(result.begin() + end_index, object(
+                    result.push_back(object(
                    {
                        {"op", "remove"},
-                        {"path", detail::concat<string_t>(path, '/', detail::to_string<string_t>(i))}
+                        {"path", detail::concat<string_t>(path, '/', detail::to_string<string_t>(j - 1))}
                    }));
-                    ++i;
                }
+                i = source.size();

                // add other remaining elements
                while (i < target.size())
@@ -223,9 +223,12 @@


 #include <array> // array
+#include <cmath> // isnan, ldexp, trunc
 #include <cstddef> // size_t
 #include <cstdint> // uint8_t
+#include <limits> // numeric_limits
 #include <string> // string
+#include <type_traits> // is_signed

 // #include <nlohmann/detail/macro_scope.hpp>
 //     __ _____ _____ _____
@@ -3283,6 +3286,68 @@ inline bool operator<(const value_t lhs, const value_t rhs) noexcept
 }
 #endif

+
+/*!
+@brief compare an integer with a floating point number without precision loss
+
+Widening the integer to the floating point type loses precision beyond the
+float's mantissa, which makes equality intransitive: both 2^63-2 and 2^63-1
+round to 2^63, so each compares equal to that float while differing from each
+other. Ordering built on that is not a strict weak ordering, so sorting such
+values, or using them as keys in an ordered container, is undefined behavior.
+
+Returns a value to be compared against zero with the original operator, which
+reproduces the exact ordering. A NaN operand is returned as is, so comparing it
+against zero keeps NaN's semantics: false for the relational operators and
+unordered for `<=>`.
+*/
+template<typename IntegerType, typename FloatType>
+FloatType compare_integer_with_float(const IntegerType i, const FloatType f) noexcept
+{
+    const auto ordered = [](int c) noexcept
+    {
+        return static_cast<FloatType>(c);
+    };
+
+    if (std::isnan(f))
+    {
+        return f;
+    }
+
+    // values of IntegerType lie in [-bound, bound) when signed and in
+    // [0, bound) when unsigned; digits excludes the sign bit, so bound is a
+    // power of two that the float represents exactly
+    const FloatType bound = std::ldexp(static_cast<FloatType>(1), std::numeric_limits<IntegerType>::digits);
+    if (f >= bound)
+    {
+        return ordered(-1);
+    }
+    if (std::is_signed<IntegerType>::value ? (f < -bound) : (f < static_cast<FloatType>(0)))
+    {
+        return ordered(1);
+    }
+
+    // f is now within the integer's range, so truncating it is exact
+    const FloatType truncated = std::trunc(f);
+    const auto as_integer = static_cast<IntegerType>(truncated);
+    if (i != as_integer)
+    {
+        return ordered(i < as_integer ? -1 : 1);
+    }
+
+    // the integer parts agree, so any fractional part decides
+    const FloatType fraction = f - truncated;
+    if (fraction > static_cast<FloatType>(0))
+    {
+        return ordered(-1);
+    }
+    if (fraction < static_cast<FloatType>(0))
+    {
+        return ordered(1);
+    }
+    return ordered(0);
+}
+
 }  // namespace detail
 NLOHMANN_JSON_NAMESPACE_END

@@ -10680,6 +10745,26 @@ inline bool little_endianness(int num = 1) noexcept
    return *reinterpret_cast<char*>(&num) == 1;
 }

+/*!
+@brief largest element count accepted for a UBJSON container of a valueless type
+
+An element of type 'Z' (null), 'T' (true) or 'F' (false) is encoded by its
+type marker alone, so an optimized container of one of those types has no
+payload at all and its declared count is the only thing that decides how much
+is allocated: `[$Z#L` followed by a large count turns some ten bytes of input
+into that many values (see #2793, which reports 35 GB and 150 seconds). Every
+other type costs at least one byte per element and is bounded by the end of
+the input.
+
+This is a sanity bound rather than a security boundary, and it is far above
+any container met in practice. @ref binary_writer falls back to the
+unoptimized encoding for longer containers, so that a value serialized by
+this library can always be read back.
+
+@sa https://github.com/nlohmann/json/issues/2793
+*/
+JSON_INLINE_VARIABLE constexpr std::size_t max_valueless_container_size = 1 << 20;
+
 ///////////////////
 // binary reader //
 ///////////////////
@@ -11618,23 +11703,21 @@ class binary_reader
    }

    /*!
-    @brief reads a CBOR string
+    @brief reads a definite-length CBOR string

-    This function first reads starting bytes to determine the expected
-    string length and then copies this number of bytes into a string.
-    Additionally, CBOR's strings with indefinite lengths are supported.
+    Reads everything @ref get_cbor_string accepts except the indefinite-length
+    form, which that function handles itself. The bytes are appended to @a
+    result, so consecutive chunks of an indefinite-length string can be read
+    into the same string.

-    @param[out] result  created string
+    @param[out] result  string the bytes are appended to

    @return whether string creation completed
-    */
-    bool get_cbor_string(string_t& result)
-    {
-        if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "string")))
-        {
-            return false;
-        }

+    @pre @a current is not EOF
+    */
+    bool get_cbor_string_chunk(string_t& result)
+    {
        switch (current)
        {
            // UTF-8 string (0x00..0x17 bytes follow)
@@ -11690,20 +11773,6 @@ class binary_reader
                return get_number(input_format_t::cbor, len) && get_string(input_format_t::cbor, len, result);
            }

-            case 0x7F: // UTF-8 string (indefinite length)
-            {
-                while (get() != 0xFF)
-                {
-                    string_t chunk;
-                    if (!get_cbor_string(chunk))
-                    {
-                        return false;
-                    }
-                    result.append(chunk);
-                }
-                return true;
-            }
-
            default:
            {
                auto last_token = get_token_string();
@@ -11714,23 +11783,82 @@ class binary_reader
    }

    /*!
-    @brief reads a CBOR byte array
+    @brief reads a CBOR string

    This function first reads starting bytes to determine the expected
-    byte array length and then copies this number of bytes into the byte array.
-    Additionally, CBOR's byte arrays with indefinite lengths are supported.
+    string length and then copies this number of bytes into a string.
+    Additionally, CBOR's strings with indefinite lengths are supported.

-    @param[out] result  created byte array
+    @param[out] result  created string
+
+    @return whether string creation completed
+    */
+    bool get_cbor_string(string_t& result)
+    {
+        // number of indefinite-length strings that have been opened and not
+        // closed yet. RFC 8949, Section 3.2.3 does not permit nesting them,
+        // but this reader has always accepted it, so the open levels are
+        // counted instead of recursed through, which overflowed the stack for
+        // an input of repeated 0x7F bytes (see #5104). Every chunk is appended
+        // to the same result, so no per-level state is needed.
+        std::size_t open = 0;
+
+        while (true)
+        {
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "string")))
+            {
+                return false;
+            }
+
+            if (current == 0x7F) // UTF-8 string (indefinite length)
+            {
+                ++open;
+                get();
+                continue;
+            }
+
+            // a break marker closes the innermost indefinite-length string;
+            // outside of one it is not a string and falls through to the error
+            if (open != 0 && current == 0xFF)
+            {
+                if (--open == 0)
+                {
+                    return true;
+                }
+                get();
+                continue;
+            }
+
+            if (JSON_HEDLEY_UNLIKELY(!get_cbor_string_chunk(result)))
+            {
+                return false;
+            }
+
+            if (open == 0)
+            {
+                return true;
+            }
+
+            get();
+        }
+    }
+
+    /*!
+    @brief reads a definite-length CBOR byte array
+
+    Reads everything @ref get_cbor_binary accepts except the indefinite-length
+    form, which that function handles itself. The bytes are appended to @a
+    result, so consecutive chunks of an indefinite-length byte array can be
+    read into the same byte array.
+
+    @param[out] result  byte array the bytes are appended to

    @return whether byte array creation completed
-    */
-    bool get_cbor_binary(binary_t& result)
-    {
-        if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "binary")))
-        {
-            return false;
-        }

+    @pre @a current is not EOF
+    */
+    bool get_cbor_binary_chunk(binary_t& result)
+    {
        switch (current)
        {
            // Binary data (0x00..0x17 bytes follow)
@@ -11790,20 +11918,6 @@ class binary_reader
                       get_binary(input_format_t::cbor, len, result);
            }

-            case 0x5F: // Binary data (indefinite length)
-            {
-                while (get() != 0xFF)
-                {
-                    binary_t chunk;
-                    if (!get_cbor_binary(chunk))
-                    {
-                        return false;
-                    }
-                    result.insert(result.end(), chunk.begin(), chunk.end());
-                }
-                return true;
-            }
-
            default:
            {
                auto last_token = get_token_string();
@@ -11813,6 +11927,63 @@ class binary_reader
        }
    }

+    /*!
+    @brief reads a CBOR byte array
+
+    This function first reads starting bytes to determine the expected
+    byte array length and then copies this number of bytes into the byte array.
+    Additionally, CBOR's byte arrays with indefinite lengths are supported.
+
+    @param[out] result  created byte array
+
+    @return whether byte array creation completed
+    */
+    bool get_cbor_binary(binary_t& result)
+    {
+        // the open indefinite-length byte arrays are counted rather than
+        // recursed through, for the reason given in @ref get_cbor_string
+        std::size_t open = 0;
+
+        while (true)
+        {
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "binary")))
+            {
+                return false;
+            }
+
+            if (current == 0x5F) // Binary data (indefinite length)
+            {
+                ++open;
+                get();
+                continue;
+            }
+
+            // a break marker closes the innermost indefinite-length byte
+            // array; outside of one it falls through to the error below
+            if (open != 0 && current == 0xFF)
+            {
+                if (--open == 0)
+                {
+                    return true;
+                }
+                get();
+                continue;
+            }
+
+            if (JSON_HEDLEY_UNLIKELY(!get_cbor_binary_chunk(result)))
+            {
+                return false;
+            }
+
+            if (open == 0)
+            {
+                return true;
+            }
+
+            get();
+        }
+    }
+
    /*!
    @brief narrow a definite CBOR array/map length to std::size_t

@@ -13013,7 +13184,12 @@ class binary_reader
    {
        result.first = npos; // size
        result.second = 0; // type
-        bool is_ndarray = false;
+        // seed the flag with the caller's context: inside an ndarray dimension
+        // vector another ndarray is not allowed, and get_ubjson_size_value()
+        // rejects it up front instead of reading it and reporting afterwards.
+        // Seeding it with `false` made every '#' of a "[#[#[..." chain descend
+        // another level, which overflowed the stack (see #5104).
+        bool is_ndarray = inside_ndarray;

        get_ignore_noop();

@@ -13046,13 +13222,11 @@ class binary_reader
            }

            const bool is_error = get_ubjson_size_value(result.first, is_ndarray);
-            if (input_format == input_format_t::bjdata && is_ndarray)
+            // an ndarray was read here only if the flag flipped; when it was
+            // seeded true, get_ubjson_size_value() already rejected the nested
+            // dimension vector
+            if (input_format == input_format_t::bjdata && is_ndarray && !inside_ndarray)
            {
-                if (inside_ndarray)
-                {
-                    return sax->parse_error(chars_read, get_token_string(), parse_error::create(112, chars_read,
-                                            exception_message(input_format, "ndarray can not be recursive", "size"), nullptr));
-                }
                result.second |= (1 << 8); // use bit 8 to indicate ndarray, all UBJSON and BJData markers should be ASCII letters
            }
            return is_error;
@@ -13061,7 +13235,7 @@ class binary_reader
        if (current == '#')
        {
            const bool is_error = get_ubjson_size_value(result.first, is_ndarray);
-            if (input_format == input_format_t::bjdata && is_ndarray)
+            if (input_format == input_format_t::bjdata && is_ndarray && !inside_ndarray)
            {
                return sax->parse_error(chars_read, get_token_string(), parse_error::create(112, chars_read,
                                        exception_message(input_format, "ndarray requires both type and size", "size"), nullptr));
@@ -13332,6 +13506,17 @@ class binary_reader

        if (size_and_type.first != npos)
        {
+            // reading an element of a valueless type consumes no input, so the
+            // declared count alone decides how much is allocated; the check is
+            // made before the start event so that no container is opened that
+            // is then abandoned. See @ref max_valueless_container_size.
+            if (JSON_HEDLEY_UNLIKELY((size_and_type.second == 'Z' || size_and_type.second == 'T' || size_and_type.second == 'F')
+                                     && size_and_type.first > max_valueless_container_size))
+            {
+                return sax->parse_error(chars_read, get_token_string(), out_of_range::create(408,
+                                        exception_message(input_format, "excessive array size", "size"), nullptr));
+            }
+
            if (JSON_HEDLEY_UNLIKELY(!sax->start_array(size_and_type.first)))
            {
                return false;
@@ -17769,7 +17954,17 @@ class binary_writer

                    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()))
+                    // 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
+                    // those, at one byte per element, so the result can be read back.
+                    // Objects are not affected: every element is preceded by its key.
+                    const bool valueless_type = (first_prefix == 'Z' || first_prefix == 'T' || first_prefix == 'F');
+                    const bool excessive_valueless = valueless_type
+                                                     && 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()))
                    {
                        prefix_required = false;
                        oa->write_character(to_char_type('$'));
@@ -18611,6 +18806,15 @@ class binary_writer
        CharType dtype = it->second;

        key = "_ArraySize_";
+        // the dimensions are written verbatim as the header length below, so a
+        // value that is not an array cannot produce a valid one: null emits 'Z'
+        // and an object emits '{', neither of which a reader accepts after '#'.
+        // Such an object is not a valid ndarray and falls back to a plain object.
+        if (!value.at(key).is_array())
+        {
+            return true;
+        }
+
        std::size_t len = (value.at(key).empty() ? 0 : 1);
        for (const auto& el : value.at(key))
        {
@@ -25057,19 +25261,19 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
    }                                                                                                    \
    else if (lhs_type == value_t::number_integer && rhs_type == value_t::number_float)                   \
    {                                                                                                    \
-        return static_cast<number_float_t>(lhs.m_data.m_value.number_integer) op rhs.m_data.m_value.number_float;      \
+        return (detail::compare_integer_with_float(lhs.m_data.m_value.number_integer, rhs.m_data.m_value.number_float)) op (static_cast<number_float_t>(0)); \
    }                                                                                                    \
    else if (lhs_type == value_t::number_float && rhs_type == value_t::number_integer)                   \
    {                                                                                                    \
-        return lhs.m_data.m_value.number_float op static_cast<number_float_t>(rhs.m_data.m_value.number_integer);      \
+        return (static_cast<number_float_t>(0)) op (detail::compare_integer_with_float(rhs.m_data.m_value.number_integer, lhs.m_data.m_value.number_float)); \
    }                                                                                                    \
    else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_float)                  \
    {                                                                                                    \
-        return static_cast<number_float_t>(lhs.m_data.m_value.number_unsigned) op rhs.m_data.m_value.number_float;     \
+        return (detail::compare_integer_with_float(lhs.m_data.m_value.number_unsigned, rhs.m_data.m_value.number_float)) op (static_cast<number_float_t>(0)); \
    }                                                                                                    \
    else if (lhs_type == value_t::number_float && rhs_type == value_t::number_unsigned)                  \
    {                                                                                                    \
-        return lhs.m_data.m_value.number_float op static_cast<number_float_t>(rhs.m_data.m_value.number_unsigned);     \
+        return (static_cast<number_float_t>(0)) op (detail::compare_integer_with_float(rhs.m_data.m_value.number_unsigned, lhs.m_data.m_value.number_float)); \
    }                                                                                                    \
    else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_integer)                \
    {                                                                                                    \
@@ -25828,7 +26032,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::forward<InputType>(i));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::cbor).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in CBOR format (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -25845,7 +26049,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::move(first), std::move(last));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::cbor).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    template<typename T>
@@ -25871,7 +26075,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::cbor).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in MessagePack format
@@ -25886,7 +26090,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::forward<InputType>(i));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::msgpack).sax_parse(input_format_t::msgpack, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in MessagePack format (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -25902,7 +26106,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::move(first), std::move(last));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::msgpack).sax_parse(input_format_t::msgpack, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    template<typename T>
@@ -25926,7 +26130,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::msgpack).sax_parse(input_format_t::msgpack, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in UBJSON format
@@ -25941,7 +26145,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::forward<InputType>(i));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::ubjson).sax_parse(input_format_t::ubjson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in UBJSON format (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -25957,7 +26161,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::move(first), std::move(last));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::ubjson).sax_parse(input_format_t::ubjson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    template<typename T>
@@ -25981,7 +26185,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::ubjson).sax_parse(input_format_t::ubjson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in BJData format
@@ -25996,7 +26200,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::forward<InputType>(i));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::bjdata).sax_parse(input_format_t::bjdata, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in BJData format (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -26012,7 +26216,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::move(first), std::move(last));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::bjdata).sax_parse(input_format_t::bjdata, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in BSON format
@@ -26027,7 +26231,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::forward<InputType>(i));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::bson).sax_parse(input_format_t::bson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    /// @brief create a JSON value from an input in BSON format (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -26043,7 +26247,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        auto ia = detail::input_adapter(std::move(first), std::move(last));
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::bson).sax_parse(input_format_t::bson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }

    template<typename T>
@@ -26067,7 +26271,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        detail::json_sax_dom_parser<basic_json, decltype(ia)> sdp(result, allow_exceptions);
        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
        const bool res = binary_reader<decltype(ia)>(std::move(ia), input_format_t::bson).sax_parse(input_format_t::bson, &sdp, strict); // cppcheck-suppress[accessMoved]
-        return res ? result : basic_json(value_t::discarded);
+        return res ? std::move(result) : basic_json(value_t::discarded);
    }
    /// @}

@@ -26482,19 +26686,17 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
                // We now reached the end of at least one array
                // in a second pass, traverse the remaining elements

-                // remove my remaining elements
-                const auto end_index = static_cast<difference_type>(result.size());
-                while (i < source.size())
+                // remove my remaining elements, highest index first; appending
+                // in that order avoids the quadratic reinsertion done before
+                for (std::size_t j = source.size(); j > i; --j)
                {
-                    // add operations in reverse order to avoid invalid
-                    // indices
-                    result.insert(result.begin() + end_index, object(
+                    result.push_back(object(
                    {
                        {"op", "remove"},
-                        {"path", detail::concat<string_t>(path, '/', detail::to_string<string_t>(i))}
+                        {"path", detail::concat<string_t>(path, '/', detail::to_string<string_t>(j - 1))}
                    }));
-                    ++i;
                }
+                i = source.size();

                // add other remaining elements
                while (i < target.size())
@@ -2751,6 +2751,31 @@ TEST_CASE("BJData")
                CHECK(json::to_bjdata(j_ok) == std::vector<uint8_t>({'[', '$', 'U', '#', '[', 'i', 2, 'i', 3, ']', 1, 2, 3, 4, 5, 6}));
                CHECK(json::from_bjdata(json::to_bjdata(j_ok), true, true) == j_ok);
            }
+
+            SECTION("ndarray whose _ArraySize_ is not an array stays as object")
+            {
+                // the shape is written verbatim as the header length, so a
+                // value that is not an array cannot produce a valid one: null
+                // would emit 'Z' and an object '{', neither of which a reader
+                // accepts after '#'. Both have to stay plain objects.
+                json const j_null = json({{"_ArrayType_", "uint8"}, {"_ArraySize_", nullptr}, {"_ArrayData_", json::array()}});
+                const auto out_null = json::to_bjdata(j_null);
+                CHECK(out_null.at(0) == '{');
+                CHECK(json::from_bjdata(out_null) == j_null);
+
+                // an object shape passes the per-entry check by iterating its
+                // values rather than dimensions, so it needs rejecting too
+                json const j_obj = json({{"_ArrayType_", "uint8"}, {"_ArraySize_", {{"a", 1}}}, {"_ArrayData_", {1}}});
+                const auto out_obj = json::to_bjdata(j_obj);
+                CHECK(out_obj.at(0) == '{');
+                CHECK(json::from_bjdata(out_obj) == j_obj);
+
+                // a scalar shape is not a dimension list either
+                json const j_num = json({{"_ArrayType_", "uint8"}, {"_ArraySize_", 1}, {"_ArrayData_", {1}}});
+                const auto out_num = json::to_bjdata(j_num);
+                CHECK(out_num.at(0) == '{');
+                CHECK(json::from_bjdata(out_num) == j_num);
+            }
        }
    }

@@ -3263,8 +3288,10 @@ TEST_CASE("BJData")
            CHECK_THROWS_WITH_AS(_ = json::from_bjdata(vR1), "[json.exception.parse_error.113] parse error at byte 6: syntax error while parsing BJData size: ndarray dimensional vector is not allowed", json::parse_error&);
            CHECK(json::from_bjdata(vR1, true, false).is_discarded());

+            // a dimension vector that opens another one is rejected where the
+            // nested '[' is read, rather than after it has been descended into
            std::vector<uint8_t> const vR2 = {'[', '$', 'i', '#', '[', '#', '[', 'i', 1, ']', ']', 1};
-            CHECK_THROWS_WITH_AS(_ = json::from_bjdata(vR2), "[json.exception.parse_error.113] parse error at byte 11: syntax error while parsing BJData size: expected length type specification (U, i, u, I, m, l, M, L) after '#'; last byte: 0x5D", json::parse_error&);
+            CHECK_THROWS_WITH_AS(_ = json::from_bjdata(vR2), "[json.exception.parse_error.113] parse error at byte 7: syntax error while parsing BJData size: ndarray dimensional vector is not allowed", json::parse_error&);
            CHECK(json::from_bjdata(vR2, true, false).is_discarded());

            std::vector<uint8_t> const vR3 = {'[', '#', '[', 'i', '2', 'i', 2, ']'};
@@ -3272,7 +3299,7 @@ TEST_CASE("BJData")
            CHECK(json::from_bjdata(vR3, true, false).is_discarded());

            std::vector<uint8_t> const vR4 = {'[', '$', 'i', '#', '[', '$', 'i', '#', '[', 'i', 1, ']', 1};
-            CHECK_THROWS_WITH_AS(_ = json::from_bjdata(vR4), "[json.exception.parse_error.110] parse error at byte 14: syntax error while parsing BJData number: unexpected end of input", json::parse_error&);
+            CHECK_THROWS_WITH_AS(_ = json::from_bjdata(vR4), "[json.exception.parse_error.113] parse error at byte 9: syntax error while parsing BJData size: ndarray dimensional vector is not allowed", json::parse_error&);
            CHECK(json::from_bjdata(vR4, true, false).is_discarded());

            std::vector<uint8_t> const vR5 = {'[', '$', 'i', '#', '[', '[', '[', ']', ']', ']'};
@@ -3280,12 +3307,25 @@ TEST_CASE("BJData")
            CHECK(json::from_bjdata(vR5, true, false).is_discarded());

            std::vector<uint8_t> const vR6 = {'[', '$', 'i', '#', '[', '$', 'i', '#', '[', 'i', '2', 'i', 2, ']'};
-            CHECK_THROWS_WITH_AS(_ = json::from_bjdata(vR6), "[json.exception.parse_error.112] parse error at byte 14: syntax error while parsing BJData size: ndarray can not be recursive", json::parse_error&);
+            CHECK_THROWS_WITH_AS(_ = json::from_bjdata(vR6), "[json.exception.parse_error.113] parse error at byte 9: syntax error while parsing BJData size: ndarray dimensional vector is not allowed", json::parse_error&);
            CHECK(json::from_bjdata(vR6, true, false).is_discarded());

            std::vector<uint8_t> const vH = {'[', 'H', '[', '#', '[', '$', 'i', '#', '[', 'i', '2', 'i', 2, ']'};
            CHECK_THROWS_WITH_AS(_ = json::from_bjdata(vH), "[json.exception.parse_error.113] parse error at byte 3: syntax error while parsing BJData size: ndarray dimensional vector is not allowed", json::parse_error&);
            CHECK(json::from_bjdata(vH, true, false).is_discarded());
+
+            // Every "#[" of this chain used to open another dimension vector
+            // and cost several stack frames before anything was rejected, so a
+            // long enough chain crashed the process (see #5104). The nested
+            // vector is refused where it is read, so the length is irrelevant.
+            std::vector<uint8_t> vRdeep = {'['};
+            for (std::size_t i = 0; i < 100000; ++i)
+            {
+                vRdeep.push_back('#');
+                vRdeep.push_back('[');
+            }
+            CHECK_THROWS_WITH_AS(_ = json::from_bjdata(vRdeep), "[json.exception.parse_error.113] parse error at byte 5: syntax error while parsing BJData size: ndarray dimensional vector is not allowed", json::parse_error&);
+            CHECK(json::from_bjdata(vRdeep, true, false).is_discarded());
        }

        SECTION("objects")
@@ -2035,6 +2035,58 @@ TEST_CASE("CBOR definite length equal to the indefinite-length sentinel")
    }
 }

+TEST_CASE("CBOR indefinite-length strings do not recurse per chunk")
+{
+    // Reading an indefinite-length string or byte array used to call itself
+    // once per chunk, so a payload of repeated 0x7F (or 0x5F) bytes exhausted
+    // the call stack before any of the input was rejected. The open levels are
+    // counted now, and the levels below prove the reader still reads the same
+    // values and reports the same errors at the same byte offsets.
+    json _;
+
+    SECTION("many open levels are reported, not crashed on")
+    {
+        const std::vector<uint8_t> input(200000, 0x7F);
+        CHECK_THROWS_WITH_AS(_ = json::from_cbor(input), "[json.exception.parse_error.110] parse error at byte 200001: syntax error while parsing CBOR string: unexpected end of input", json::parse_error&);
+        CHECK(json::from_cbor(input, true, false).is_discarded());
+    }
+
+    SECTION("many open levels are reported, not crashed on (binary)")
+    {
+        const std::vector<uint8_t> input(200000, 0x5F);
+        CHECK_THROWS_WITH_AS(_ = json::from_cbor(input), "[json.exception.parse_error.110] parse error at byte 200001: syntax error while parsing CBOR binary: unexpected end of input", json::parse_error&);
+        CHECK(json::from_cbor(input, true, false).is_discarded());
+    }
+
+    SECTION("chunks are still concatenated")
+    {
+        CHECK(json::from_cbor(std::vector<uint8_t>({0x7F, 0xFF})) == json(""));
+        CHECK(json::from_cbor(std::vector<uint8_t>({0x7F, 0x61, 0x61, 0xFF})) == json("a"));
+        // nested indefinite-length strings are concatenated across levels
+        CHECK(json::from_cbor(std::vector<uint8_t>({0x7F, 0x7F, 0x61, 0x61, 0xFF, 0x61, 0x62, 0xFF})) == json("ab"));
+        CHECK(json::from_cbor(std::vector<uint8_t>({0x7F, 0x7F, 0x7F, 0x61, 0x7A, 0xFF, 0xFF, 0xFF})) == json("z"));
+        CHECK(json::from_cbor(std::vector<uint8_t>({0xA1, 0x7F, 0x61, 0x61, 0xFF, 0x01})) == json({{"a", 1}}));
+    }
+
+    SECTION("chunks are still concatenated (binary)")
+    {
+        CHECK(json::from_cbor(std::vector<uint8_t>({0x5F, 0x41, 0x61, 0xFF})) == json::binary({0x61}));
+        CHECK(json::from_cbor(std::vector<uint8_t>({0x5F, 0x5F, 0x41, 0x61, 0xFF, 0x41, 0x62, 0xFF})) == json::binary({0x61, 0x62}));
+    }
+
+    SECTION("a chunk that is not a string is still rejected")
+    {
+        CHECK_THROWS_WITH_AS(_ = json::from_cbor(std::vector<uint8_t>({0x7F, 0x7F, 0x00})), "[json.exception.parse_error.113] parse error at byte 3: syntax error while parsing CBOR string: expected length specification (0x60-0x7B) or indefinite string type (0x7F); last byte: 0x00", json::parse_error&);
+        CHECK_THROWS_WITH_AS(_ = json::from_cbor(std::vector<uint8_t>({0x5F, 0x5F, 0x00})), "[json.exception.parse_error.113] parse error at byte 3: syntax error while parsing CBOR binary: expected length specification (0x40-0x5B) or indefinite binary array type (0x5F); last byte: 0x00", json::parse_error&);
+    }
+
+    SECTION("a break marker outside an indefinite-length string is not a string")
+    {
+        // 0xFF only closes a string that was opened; on its own it is not one
+        CHECK_THROWS_WITH_AS(_ = json::from_cbor(std::vector<uint8_t>({0xA1, 0xFF, 0x01})), "[json.exception.parse_error.113] parse error at byte 2: syntax error while parsing CBOR string: expected length specification (0x60-0x7B) or indefinite string type (0x7F); last byte: 0xFF", json::parse_error&);
+    }
+}
+
 TEST_CASE("CBOR roundtrips" * doctest::skip())
 {
    SECTION("input from flynn")
@@ -326,6 +326,57 @@ TEST_CASE("lexicographical comparison operators")
 #endif
        }

+        SECTION("integer/float mixed comparison is exact")
+        {
+            // Widening the integer to a double loses precision past the
+            // mantissa, so 2^63-2 and 2^63-1 both used to compare equal to the
+            // double 2^63 while differing from each other. That makes equality
+            // intransitive and the ordering not a strict weak ordering.
+            const json below_two_63 = static_cast<std::int64_t>(9223372036854775806LL);
+            const json max_int64 = (std::numeric_limits<std::int64_t>::max)();
+            const json two_63 = 9223372036854775808.0;
+
+            CHECK_FALSE(below_two_63 == two_63);
+            CHECK_FALSE(max_int64 == two_63);
+            CHECK(below_two_63 != max_int64);
+            CHECK(below_two_63 < max_int64);
+            CHECK(below_two_63 < two_63);
+            CHECK(max_int64 < two_63);
+            CHECK(two_63 > max_int64);
+            CHECK_FALSE(two_63 < max_int64);
+
+            // the same past the unsigned range
+            const json max_uint64 = (std::numeric_limits<std::uint64_t>::max)();
+            const json two_64 = 18446744073709551616.0;
+            CHECK_FALSE(max_uint64 == two_64);
+            CHECK(max_uint64 < two_64);
+            CHECK(two_64 > max_uint64);
+
+            // values a double represents exactly still compare equal
+            CHECK(json(1) == json(1.0));
+            CHECK(json(1u) == json(1.0));
+            CHECK(json(-3) == json(-3.0));
+            CHECK(json(1) < json(1.5));
+            CHECK(json(1.5) < json(2));
+            CHECK(json(2) > json(1.5));
+
+            // a NaN operand stays unordered against either integer kind
+            CHECK_FALSE(json(1) == json(nan));
+            CHECK_FALSE(json(1) < json(nan));
+            CHECK_FALSE(json(nan) < json(1));
+            CHECK_FALSE(json(1u) == json(nan));
+
+#if JSON_HAS_THREE_WAY_COMPARISON
+            // JSON_HAS_CPP_20 (do not remove; see note at top of file)
+            CHECK((max_int64 <=> two_63) == std::partial_ordering::less); // *NOPAD*
+            CHECK((two_63 <=> max_int64) == std::partial_ordering::greater); // *NOPAD*
+            CHECK((below_two_63 <=> max_int64) == std::partial_ordering::less); // *NOPAD*
+            CHECK((max_uint64 <=> two_64) == std::partial_ordering::less); // *NOPAD*
+            CHECK((json(1) <=> json(1.0)) == std::partial_ordering::equivalent); // *NOPAD*
+            CHECK((json(1) <=> json(nan)) == std::partial_ordering::unordered); // *NOPAD*
+#endif
+        }
+
        SECTION("compares unordered")
        {
            std::vector<std::vector<bool>> expected =
@@ -1388,3 +1388,84 @@ TEST_CASE("JSON patch - add to a primitive parent (regression #4292)")
        CHECK_THROWS_AS(doc.patch(patch), json::out_of_range&);
    }
 }
+
+TEST_CASE("JSON patch - diff emits array removals in descending index order")
+{
+    SECTION("array shrunk to empty")
+    {
+        json const source = {0, 1, 2, 3, 4};
+        json const target = json::array();
+
+        json const patch = json::diff(source, target);
+
+        json const expected = R"(
+            [
+                {"op": "remove", "path": "/4"},
+                {"op": "remove", "path": "/3"},
+                {"op": "remove", "path": "/2"},
+                {"op": "remove", "path": "/1"},
+                {"op": "remove", "path": "/0"}
+            ]
+        )"_json;
+
+        CHECK(patch == expected);
+        CHECK(source.patch(patch) == target);
+    }
+
+    SECTION("array partially shrunk, after a replacement at a common index")
+    {
+        json const source = {0, 1, 2, 3, 4};
+        json const target = {0, 9};
+
+        json const patch = json::diff(source, target);
+
+        // the replacement comes first, then the removals, highest index first
+        json const expected = R"(
+            [
+                {"op": "replace", "path": "/1", "value": 9},
+                {"op": "remove", "path": "/4"},
+                {"op": "remove", "path": "/3"},
+                {"op": "remove", "path": "/2"}
+            ]
+        )"_json;
+
+        CHECK(patch == expected);
+        CHECK(source.patch(patch) == target);
+    }
+
+    SECTION("nested array shrunk")
+    {
+        json const source = {{"a", {0, 1, 2}}};
+        json const target = {{"a", json::array()}};
+
+        json const patch = json::diff(source, target);
+
+        json const expected = R"(
+            [
+                {"op": "remove", "path": "/a/2"},
+                {"op": "remove", "path": "/a/1"},
+                {"op": "remove", "path": "/a/0"}
+            ]
+        )"_json;
+
+        CHECK(patch == expected);
+        CHECK(source.patch(patch) == target);
+    }
+
+    SECTION("many removals still round-trip")
+    {
+        json source = json::array();
+        for (int i = 0; i < 1000; ++i)
+        {
+            source.push_back(i);
+        }
+        json const target = json::array();
+
+        json const patch = json::diff(source, target);
+
+        CHECK(patch.size() == 1000);
+        CHECK(patch.front().at("path") == "/999");
+        CHECK(patch.back().at("path") == "/0");
+        CHECK(source.patch(patch) == target);
+    }
+}
@@ -2149,6 +2149,61 @@ TEST_CASE("UBJSON")
    }
 }

+TEST_CASE("UBJSON optimized arrays of a valueless type are bounded")
+{
+    // An element of type 'Z', 'T' or 'F' is encoded by its marker alone, so an
+    // optimized array of one of those has no payload and the declared count is
+    // the only thing deciding how much is allocated. Ten bytes used to produce
+    // billions of values (#2793); every other type costs at least one byte per
+    // element and is bounded by the end of the input.
+    json _;
+
+    SECTION("an excessive count is rejected")
+    {
+        // 'l' is a big-endian int32: 0x7FFFFFFF elements, about 34 GB of value
+        for (const auto marker :
+                {'Z', 'T', 'F'
+                })
+        {
+            const std::vector<uint8_t> input = {'[', '$', static_cast<uint8_t>(marker), '#', 'l', 0x7F, 0xFF, 0xFF, 0xFF};
+            CHECK_THROWS_WITH_AS(_ = json::from_ubjson(input), "[json.exception.out_of_range.408] syntax error while parsing UBJSON size: excessive array size", json::out_of_range&);
+            CHECK(json::from_ubjson(input, true, false).is_discarded());
+        }
+    }
+
+    SECTION("ordinary counts are unaffected")
+    {
+        CHECK(json::from_ubjson(std::vector<uint8_t>({'[', '$', 'Z', '#', 'i', 3})) == json({nullptr, nullptr, nullptr}));
+        CHECK(json::from_ubjson(std::vector<uint8_t>({'[', '$', 'T', '#', 'i', 2})) == json({true, true}));
+        CHECK(json::from_ubjson(std::vector<uint8_t>({'[', '$', 'F', '#', 'i', 2})) == json({false, false}));
+        // 'N' is a no-op rather than a value, and still yields an empty array
+        CHECK(json::from_ubjson(std::vector<uint8_t>({'[', '$', 'N', '#', 'i', 2})) == json::array());
+    }
+
+    SECTION("a type with a payload is unaffected")
+    {
+        // the same count for 'U' is bounded by the end of the input instead
+        const std::vector<uint8_t> input = {'[', '$', 'U', '#', 'l', 0x7F, 0xFF, 0xFF, 0xFF};
+        CHECK_THROWS_AS(_ = json::from_ubjson(input), json::parse_error&);
+    }
+
+    SECTION("the writer stays within what the reader accepts")
+    {
+        // below the limit the optimized form is used and is tiny; above it the
+        // writer falls back so that the result can still be read back
+        json const at_limit(1048576, nullptr);
+        const auto v_at_limit = json::to_ubjson(at_limit, true, true);
+        CHECK(v_at_limit.size() == 9);
+        CHECK(v_at_limit.at(1) == '$');
+        CHECK(json::from_ubjson(v_at_limit) == at_limit);
+
+        json const above_limit(1048577, nullptr);
+        const auto v_above_limit = json::to_ubjson(above_limit, true, true);
+        CHECK(v_above_limit.at(1) != '$');
+        CHECK(json::from_ubjson(v_above_limit) == above_limit);
+    }
+}
+
 TEST_CASE("Universal Binary JSON Specification Examples 1")
 {
    SECTION("Null Value")
Author	SHA1	Message	Date
Niels Lohmann	d89acce09a	Bound UBJSON optimized arrays of a valueless type An element of type 'Z' (null), 'T' (true) or 'F' (false) is encoded by its type marker alone, so an optimized UBJSON array of one of those has no payload: reading an element consumes no input at all. Its declared count is therefore the only thing that decides how much is allocated, and nothing bounded it. "[$Z#l" and a four-byte count is nine bytes of input describing two billion values; #2793 reports 35 GB and 150 seconds from ten bytes, and OSS-Fuzz has an out-of-memory and a timeout report for the same shape. Every other type costs at least one byte per element, so the end of the input bounds it. 'N' (no-op) is already skipped rather than stored. Objects are not affected either: each element is preceded by its key, which costs bytes. And BJData already refuses these markers as an optimized type, so this is a plain UBJSON matter. Reject a count above 1,048,576 elements for those three types with out_of_range.408, the code this reader already uses for a declared size it will not honour. The check runs before the SAX start event, so no container is opened and then abandoned. Rejecting on the read side alone would break the guarantee that anything to_ubjson() writes can be read back, and would trip the round-trip assertion in fuzzer-parse_ubjson.cpp. So the writer falls back to the unoptimized encoding, one byte per element, for arrays of these types above the same limit. Its decision depends only on the array's size, which is identical for a value and for anything parsed back from it, so the round trip is stable. No existing test changes: the largest such count in the test suite is 65,793. The excessive-size test that already used this shape still passes, now rejected a little earlier than by the max_size() check it used to reach. Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-09-06 18:12:41 +02:00
Niels Lohmann	bcd5af62c7	Reject a nested BJData ndarray dimension vector where it is read get_ubjson_size_type() takes an inside_ndarray parameter saying whether it is being called for an ndarray's dimension vector, where another ndarray is not allowed. It then seeded the flag it passes down to get_ubjson_size_value() with `false` rather than with that parameter, and only consulted inside_ndarray afterwards, on the '$' branch. So on the '#' branch nothing stopped the descent: every "#[" pair of an input like "[" followed by "#[#[#[..." opened another dimension vector, several native stack frames deeper each time, and the recursion was only reported on the way back out. 100,000 pairs crash the process. This is #5104 again, in a path that has nothing to do with containers. Seed the flag with inside_ndarray, which is what get_ubjson_size_value() documents it wants: "for input, `true` means already inside an ndarray vector or ndarray dimension is not allowed". The nested '[' is then refused where it is read, so the length of the chain no longer matters. Both post-checks gain `&& !inside_ndarray`, because an ndarray was found here only if the flag flipped -- get_ubjson_size_value() only ever returns `true` when its initial value was `false`, as its documentation says. With that, the "ndarray can not be recursive" branch is unreachable: a recursive ndarray is now caught one level earlier, and reported as "ndarray dimensional vector is not allowed" like every other nested dimension vector. Three existing expectations move accordingly (vR2, vR4, vR6). All three now fail earlier, and all three now report the same error that vR1, vR5 and vH already reported for the same shape, which is the more consistent outcome. Everything else is unchanged: valid 1D and 2D ndarrays, optimized containers and plain arrays produce identical results, and unit-ubjson is untouched. Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-09-06 18:12:40 +02:00
Niels Lohmann	25a4333a31	Stop CBOR indefinite-length strings from recursing per chunk get_cbor_string() and get_cbor_binary() handled the indefinite-length forms (0x7F and 0x5F) by calling themselves once per chunk. Each chunk therefore cost a native stack frame, and since a chunk may itself be an indefinite- length string, an input of repeated 0x7F bytes reached one frame per input byte: 200,000 of them crash the process with SIGSEGV before a single byte is rejected. This is the same defect as #5104, in a path the container-level work does not touch. Count the open levels instead of recursing through them. That is enough here because every chunk is appended to the same result -- get_bytes() writes at result.size() -- so there is no per-level state to keep. The temporary chunk string and its copy into the result go away with the recursion. The definite-length cases move to get_cbor_string_chunk() and get_cbor_binary_chunk() unchanged, including their error messages, which still name 0x7F and 0x5F because those are handled one level up. Behaviour is unchanged. Comparing against develop over the interesting byte sequences -- empty, single-chunk, nested, over-closed and truncated forms, both strings and byte arrays, and an indefinite-length map key -- produces identical values, error codes, messages and byte offsets. The 200,000-level input now reports parse_error.110 at byte 200001 instead of crashing. Note that nesting these is not valid CBOR: RFC 8949, Section 3.2.3 forbids it. This does not change that either way -- it has always been accepted, and rejecting it is a separate decision (#5317, #5325). Should it be rejected later, that is now one condition on the level counter rather than a change to the control flow. Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-09-06 18:12:39 +02:00
Niels Lohmann	c3219fdc30	Return the parsed value by move from from_cbor() and friends The binary entry points end with return res ? result : basic_json(value_t::discarded); The condition operator's second operand is an lvalue, so this is not a case where the return value can be elided or implicitly moved from: every successful from_cbor(), from_msgpack(), from_ubjson(), from_bjdata() and from_bson() call deep-copies the value it just parsed, and then destroys the original. The copy is not cheap, and it is not incidental: basic_json's copy constructor walks the whole value. Parsing a 2 MB CBOR document with 60,000 objects, median of 25 runs, clang 17 -O3: from_cbor 26.99 ms -> 14.65 ms from_msgpack 26.82 ms -> 14.82 ms Moving instead of copying is the entire change; the parsed value is not used again after the return expression is evaluated. There is a second reason to prefer the move. The copy constructor recurses once per nesting level, so the copy is also a stack-overflow path on the return side, on a value the reader has already accepted. That is currently masked because the readers themselves recurse and overflow first (#5104), but it has to be fixed for making them iterative to have any effect. Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-09-06 16:41:35 +02:00
Qatadaha Bin Matloob	09b6b6b5ba	Fix to_bjdata() emitting unparsable output when _ArraySize_ is not an array (#5455 ) * Fix to_bjdata() emitting unparsable output when _ArraySize_ is not an array write_bjdata_ndarray() never checked that _ArraySize_ is an array. The shape is written verbatim as the header length, so a null shape emitted 'Z' and an object shape emitted '{' after the '#', neither of which from_bjdata() accepts, and the round-trip guarantee in the BJData docs was broken. Both slipped through the existing validation: for null, empty() is true so the element count starts at 0 and the per-dimension loop never runs, and for an object the loop walks its values, which can satisfy the non-negative integer check. When _ArrayData_ then matched that count, the writer took the ndarray path. Require the shape to be an array, so anything else falls back to a plain object encoding that round-trips, as the fallback rule in the docs already specifies. Signed-off-by: qatcod <79017227+qatcod@users.noreply.github.com> * Document that _ArraySize_ must be an array in the ndarray requirements The list at bjdata.md is the exhaustive set of conditions for the ndarray encoding, but it only implied this one through 'every entry of'. Signed-off-by: qatcod <79017227+qatcod@users.noreply.github.com> --------- Signed-off-by: qatcod <79017227+qatcod@users.noreply.github.com>	2026-09-05 20:30:22 +02:00
Niels Lohmann	dd26d3ff07	docs: add 41 customers and rebuild the overview image (#5463 ) * docs: add 89 customers and sort all sections Extend the customers page from 136 to 225 entries. New entries were found by searching vendor open-source notices and by inspecting dependency manifests in public repositories. Every added entry was verified to link either to a page that credits the library, or to an open source repository where its use is directly visible (a vendored copy, a build manifest, or an include). Candidates whose only evidence was a transitive dependency (via ICU or KDDockWidgets), packaging metadata, or an unused vendored file were dropped rather than listed. Also sort every section alphabetically ignoring case, and keep the Peregrine lunar lander first under "Space Exploration". Signed-off-by: Niels Lohmann <mail@nlohmann.me> * docs: add 41 customers and rebuild the overview image The list was assembled by searching for names one at a time, which is why it had plateaued. These additions come from enumerating instead: a public code search returning every repository that references the library, and the reverse build-dependencies of the Debian and Ubuntu source indices. Every addition cites a first-party call site rather than a repository root, and every evidence URL was checked to resolve. Several prominent candidates were rejected on exactly that test: - Node.js reaches the library only through deps/icu-small, the inherited attribution ICU carries into everything that ships it. - Visual Studio Code's only hits are copies of the library's own headers used as sample C++ in the Copilot extension's test fixtures. - OSS-Fuzz fuzzes the library rather than calling it; projects/json/ is the library's own OSS-Fuzz integration. - libuv matched only its AUTHORS file, which lists a contributor by name. - simdjson and LLVM matched only benchmarks and a mangled-symbol test. Two entries were already present under a different name and are merged rather than duplicated: PrestoDB, and DB Browser for SQLite under its repository name. GitHub CodeQL's citation moves from the repository root to shared/cpp/Diagnostics.h, which includes the header directly. The image is rebuilt with 236 logos over 17 rows. It is smaller than the one it replaces, 1.02 MB against 1.35 MB, despite carrying 100 more marks. Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-09-05 11:27:57 +02:00
Avionic Harshit	3c875683a5	Make diff() linear when an array shrinks (#5461 ) Signed-off-by: avionicharshit-byte <harshitavionic@gmail.com>	2026-09-05 09:52:53 +02:00
Qatadaha Bin Matloob	137a40b9aa	Compare integers with floats exactly instead of widening the integer (#5459 ) The mixed number arms of JSON_IMPLEMENT_OPERATOR cast the integer to number_float_t before comparing. Past the float's mantissa that cast is lossy: 2^63-2 and 2^63-1 both round to 2^63, so each compares equal to that float while differing from each other. Equality is therefore intransitive and the ordering is not a strict weak ordering, which makes std::sort over such values, or using them as keys in std::set or std::map, undefined behavior. Compare the two exactly instead. The integer's range is a power of two the float represents exactly, so a float outside it is ordered by magnitude alone; inside it, truncating the float is exact, and the integer parts and then any fractional part decide. The helper hands back a pair whose comparison with the original operator reproduces that ordering, which keeps every operator's return type as it was, including partial_ordering for the spaceship. A NaN operand is returned in both members, so NaN stays false for the relational operators and unordered for <=>. Values a float represents exactly still compare equal, so json(1) == json(1.0) is unchanged. Signed-off-by: qatcod <79017227+qatcod@users.noreply.github.com>	2026-09-05 09:43:07 +02:00
dependabot[bot]	19386dd14a	Bump the codeql-action group with 4 updates (#5454 ) Bumps the codeql-action group with 4 updates: [github/codeql-action/init](https://github.com/github/codeql-action), [github/codeql-action/autobuild](https://github.com/github/codeql-action), [github/codeql-action/analyze](https://github.com/github/codeql-action) and [github/codeql-action/upload-sarif](https://github.com/github/codeql-action). Updates `github/codeql-action/init` from 4.37.8 to 4.37.9 - [Release notes](https://github.com/github/codeql-action/releases) - [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md) - [Commits](https://github.com/github/codeql-action/compare/db488ddef3bf6cb639b32c2e9a7c0a7ea8271d28...cdf488f595d80d6e07e03d4674febd5ab45fa938) Updates `github/codeql-action/autobuild` from 4.37.8 to 4.37.9 - [Release notes](https://github.com/github/codeql-action/releases) - [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md) - [Commits](https://github.com/github/codeql-action/compare/db488ddef3bf6cb639b32c2e9a7c0a7ea8271d28...cdf488f595d80d6e07e03d4674febd5ab45fa938) Updates `github/codeql-action/analyze` from 4.37.8 to 4.37.9 - [Release notes](https://github.com/github/codeql-action/releases) - [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md) - [Commits](https://github.com/github/codeql-action/compare/db488ddef3bf6cb639b32c2e9a7c0a7ea8271d28...cdf488f595d80d6e07e03d4674febd5ab45fa938) Updates `github/codeql-action/upload-sarif` from 4.37.8 to 4.37.9 - [Release notes](https://github.com/github/codeql-action/releases) - [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md) - [Commits](https://github.com/github/codeql-action/compare/db488ddef3bf6cb639b32c2e9a7c0a7ea8271d28...cdf488f595d80d6e07e03d4674febd5ab45fa938) --- updated-dependencies: - dependency-name: github/codeql-action/init dependency-version: 4.37.9 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: codeql-action - dependency-name: github/codeql-action/autobuild dependency-version: 4.37.9 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: codeql-action - dependency-name: github/codeql-action/analyze dependency-version: 4.37.9 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: codeql-action - dependency-name: github/codeql-action/upload-sarif dependency-version: 4.37.9 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: codeql-action ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>	2026-09-02 22:34:12 +02:00