Skip to content

Regenerate CI workflows with xmsconan 2.18.0 - #130

Merged
gagelarsen merged 1 commit into
masterfrom
bump/xmsconan-2.18.0
Aug 13, 2026
Merged

Regenerate CI workflows with xmsconan 2.18.0#130
gagelarsen merged 1 commit into
masterfrom
bump/xmsconan-2.18.0

Conversation

@gagelarsen

Copy link
Copy Markdown
Member

Bumps generated CI onto xmsconan 2.18.0 (The VS2019 Bridge Release). Five lines, all pins.

Why this is required, not housekeeping

XmsConan2File.export() copies xms_conan2_file.py into the export folder, so its contents feed the recipe revision hash. 2.18.0 changes that file by 192 lines — the msvc 192 fork (_is_vs2019, vs2019_requirements, vs2019_dependency_overrides).

That means a 7.0.x built by CI on the 2.16.0 pin and the same 7.0.x built locally for VS2019 on 2.18.0 export different recipes and resolve to different rrevs. aquaveo-stable and aquaveo-vs2019 would be silently out of step on a version that reads as one release — precisely the failure the VS2019 bridge exists to avoid. Both sides have to generate from the same xmsconan.

Blast radius on the VS2022 side: none

Diffed the generated output against a real xmsconan==2.16.0 install (not the editable one):

File 2.16.0 vs 2.18.0
conanfile.py identical
CMakeLists.txt identical
pytest.ini identical
build.py exit 1 when builder.upload() reports errors, instead of failing silently. Not exported.
xms_conan2_file.py 192 lines — the vs2019 fork. This is the intended rrev move.

conanfile.py gains a vs2019_dependency_overrides attribute only when build.toml declares that table; this repo doesn't, so it regenerates byte-identical.

On any toolchain that is not msvc 192 the dependency graph is unchanged. 2.18.0 refactored the inline self.requires("boost/1.86.0") / self.requires("zlib/1.3.1") calls into an iteration over default_requirements, which holds those same two references in the same order.

This should be the last regeneration commit

2.17.0 floated generated CI from == onto >= with --upgrade, so a future xmsconan reaches CI on its next run rather than waiting on a regenerate-and-commit pass. The flake job picks up --upgrade here, which it lacked under the == pin.

Note Coverage.yaml was already floating (>=2.16.0); only its floor moves.

🤖 Generated with Claude Code

CI has to generate the recipe with the same xmsconan the manual VS2019 track
uses, or the two remotes end up carrying different recipe revisions for the
same version.

XmsConan2File.export() copies xms_conan2_file.py into the export folder, so its
contents feed the recipe revision hash. 2.18.0 changes that file by 192 lines
(the msvc 192 fork: _is_vs2019, vs2019_requirements, vs2019_dependency_
overrides). A 7.0.x built by CI on the 2.16.0 pin and the same 7.0.x built
locally for VS2019 on 2.18.0 would therefore export different recipes and
resolve to different rrevs -- aquaveo-stable and aquaveo-vs2019 silently out of
step on a version that reads as one release.

Nothing else in the generated output moves. Diffed against a real 2.16.0
install: conanfile.py, CMakeLists.txt and pytest.ini are byte-identical
(conanfile.py gains a vs2019_dependency_overrides attribute only when
build.toml declares that table, and this one does not). build.py gains an exit
1 when builder.upload() reports errors instead of failing silently, and is not
exported. On any toolchain that is not msvc 192 the dependency graph is
unchanged: 2.18.0 refactored the inline self.requires("boost/1.86.0") /
self.requires("zlib/1.3.1") calls into an iteration over default_requirements
holding those same references in the same order.

The pins also stop being pins. 2.17.0 floated generated CI onto
>= with --upgrade, so this is the last regeneration commit this repo needs --
a later xmsconan reaches CI on its next run. The flake job picks up --upgrade
here, which it lacked under the == pin.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
@gagelarsen
gagelarsen merged commit 2a5bd74 into master Aug 13, 2026
24 checks passed
@gagelarsen
gagelarsen deleted the bump/xmsconan-2.18.0 branch August 13, 2026 21:43
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant