On Wed, Sep 17, 2025 at 12:01:28PM +0200, Bill Allombert wrote: > On Wed, Sep 17, 2025 at 08:33:50AM +0000, Gianfranco Costamagna wrote: > > export CFLAGS=-DNLOHMANN_JSON_NAMESPACE_NO_VERSION=1Hello, new idea, what if > > I upload 3.12 with abi stick to 3.11, such as not changing the namespace > > anymore? > > #define NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c) json_abi ## a ## b ## c > > It is unreasonnable to change the ABI without upstream backing. We should not > create Debian-specific ABI. > > I encourage you to discuss the ABI issue with nlohmann-json3 upstream to find > an agreement for future version of nlohmann-json3-dev, but now the urgency is > to restore nlohmann-json3-dev to 3.11.3 to fix all the RC bugs that this > cause, > and check every libraries for the wrong ABI and ask for a binNMU.
So I used codesearch.debian.net and I found that quite a few packages embed their own copy of nlohmann/json.hpp instead of build-depending on this package: avogadrolibs, babeltrace2, blender, boost1.88, codeblocks, dd-opentracing-cpp, firefox, firefox-esr, geos, gfxreconstruct, gtkwave, guitarix, icu, kicad, libkysdk-base, llama.cpp, mariadb, mkvtoolnix, mozjs140, mumble, newsboat, ns3, nvidia-cudnn-frontend, openttd, pbcopper, picotool, poedit, qtcreator, satdump, snapcast, stella, ticcutils, tracy, warzone2100 For example: https://sources.debian.org/sha256/?checksum=9bea4c8066ef4a1c206b2be5a36302f8926f7fdc6087af5d20b417d0cf103ea6&package= Checksum: 9bea4c8066ef4a1c206b2be5a36302f8926f7fdc6087af5d20b417d0cf103ea6 (page 1) 22 results: blender/4.3.2+dfsg-2/extern/json/include/json.hpp boost1.88/1.88.0-1/tools/build/src/engine/ext_nlohmann_json.hpp boost1.88/1.88.0-1.3/tools/build/src/engine/ext_nlohmann_json.hpp boost1.88/1.88.0-1.4/tools/build/src/engine/ext_nlohmann_json.hpp f3d/3.1.0+dfsg-1/external/nlohmann_json/nlohmann/json.hpp kicad/9.0.2+dfsg-1/thirdparty/nlohmann_json/nlohmann/json.hpp kicad/9.0.3+dfsg-1~bpo12+1/thirdparty/nlohmann_json/nlohmann/json.hpp kicad/9.0.4+dfsg-1/thirdparty/nlohmann_json/nlohmann/json.hpp kicad/9.0.4+dfsg-1~bpo13+1/thirdparty/nlohmann_json/nlohmann/json.hpp mariadb/1:11.8.3-0+deb13u1/storage/columnstore/columnstore/utils/json/json.hpp mariadb/1:11.8.3-1/storage/columnstore/columnstore/utils/json/json.hpp mumble/1.5.735-5/3rdparty/nlohmann_json/single_include/nlohmann/json.hpp mumble/1.5.735-6/3rdparty/nlohmann_json/single_include/nlohmann/json.hpp newsboat/2.36-1.1/3rd-party/json.hpp nlohmann-json3/3.11.3-2.1/single_include/nlohmann/json.hpp nlohmann-json3/3.12.0.really.3.11.3-0.1/single_include/nlohmann/json.hpp openttd/14.1-1/src/3rdparty/nlohmann/json.hpp snapcast/0.28.0-1~bpo12+1/common/json.hpp snapcast/0.30.0-1~bpo12+1/common/json.hpp snapcast/0.31.0-1/common/json.hpp tracy/0.11.0+ds-1/import-chrome/src/json.hpp warzone2100/4.5.5deb-1/3rdparty/json/include/nlohmann/json.hpp This is all byt for byte copy of nlohmann-json3/3.11.3-2.1/single_include/nlohmann/json.hpp I will reopen the RC bug to prevent 3.12.0 to reach testing. Cheers, -- Bill. <[email protected]> Imagine a large red swirl here.
