Summary: | media-libs/assimp-5.2.5-r1 fails AssimpAPITest_aiVector3D.aiVector3RotateByQuaternionTest test with LTO | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Paolo Pedroni <paolo.pedroni> |
Component: | Current packages | Assignee: | Bernd <waebbl-gentoo> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | proxy-maint |
Priority: | Normal | Keywords: | PullRequest |
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
See Also: | https://github.com/gentoo/gentoo/pull/34346 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | |||
Bug Blocks: | 618550, 915000 | ||
Attachments: | assimp-5.2.5-r1:20230612-115707.log.gz |
Description
Paolo Pedroni
2023-06-12 12:15:24 UTC
This is a pretty suspicious failure... Does it get better with different flags? -O2 -march=znver2 works for me -O2 -march=znver2 -ftree-vectorize -flto=8 fails.. (In reply to Sam James from comment #2) > -O2 -march=znver2 works for me > -O2 -march=znver2 -ftree-vectorize -flto=8 fails.. with -O2 -march=znver2 -flto, I get: [ FAILED ] 2 tests, listed below: [ FAILED ] AssimpAPITest_aiVector3D.aiVector3RotateByQuaternionTest [ FAILED ] AssimpAPITest_aiQuaternion.aiQuaternionMultiplyTest not sure if the 1 extra failure is just a flake or if no -ftree-vectorize makes it worse Would be interesting to know if master is any better, I see a lot of correctness fixes. Just for reference since GCC and Clang handle LTO differently, that Clang and -flto passes all tests on my LLVM machine. The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=054f5ce87ab5e96a7224e2e396f897472d0d8f62 commit 054f5ce87ab5e96a7224e2e396f897472d0d8f62 Author: Paul Zander <negril.nx+gentoo@gmail.com> AuthorDate: 2023-12-15 16:50:01 +0000 Commit: Sam James <sam@gentoo.org> CommitDate: 2024-01-20 12:26:30 +0000 media-libs/assimp: add 5.2.5-r2, add 5.3.1 Rework tests to make use of gtest via ctest. Some tests generate random input data which might fail on fma-enabled CPUs due to smaller rounding errors on fused operations, so we repeat those tests until they succeed. Closes: https://bugs.gentoo.org/908403 Closes: https://bugs.gentoo.org/910362 Signed-off-by: Paul Zander <negril.nx+gentoo@gmail.com> Closes: https://github.com/gentoo/gentoo/pull/34346 Signed-off-by: Sam James <sam@gentoo.org> media-libs/assimp/Manifest | 1 + media-libs/assimp/assimp-5.2.5-r2.ebuild | 94 ++++++++++++++++ media-libs/assimp/assimp-5.3.1.ebuild | 125 +++++++++++++++++++++ .../files/assimp-5.2.5-float-comparison.patch | 84 ++++++++++++++ media-libs/assimp/files/assimp-5.2.5-gtest.patch | 30 +++++ 5 files changed, 334 insertions(+) |