https://blogs.gentoo.org/ago/2020/07/04/gentoo-tinderbox/ Issue: sci-mathematics/singular-4.2.0_p1 does not respect CFLAGS. Discovered on: amd64 (internal ref: ci) NOTE: To reproduce this issue you may want to add "-frecord-gcc-switches" to CFLAGS/CXXFLAGS/FFLAGS/FCFLAGS. If this is not something about c/c++ context see this bug as an hint to hide the QA warning (with QA_FLAGS_IGNORED) where is not possible to respect CFLAGS.
Created attachment 693117 [details] build.log.xz build log and emerge --info (compressed because it exceeds attachment limit, use 'xzless' to read it)
Thanks for report. To think I worked on this for that much time and it fails on this.
OK, this is very telling * QA Notice: Files built without respecting CFLAGS have been detected * Please include the following list of files in your report: * /usr/libexec/singular/MOD/python_module.so this shouldn't be built for several reasons in the first place.
OK, this is fun. Apart from the old AX_PYTHON_* macros of old there is a new SING_CHECK_PYTHON_MODULE in m4/options.m4. The first set of macros define PYTHON_USE which default to "false". The new macro define defines ENABLE_PYTHON_MODULE and the default is "no". But neither variables have an impact on whether Singular/dyn_modules/{python,pyobject}/* are built or not. In fact they are not used anywhere outside the macros themselves and the top configure file. This has been fixed on the master branch as part of all the PRs Matthias Koeppe from sage has sent. It would be rather hard to include everything. If we want to fix this version in particular we may need precise surgery instead.
I have a confession to make: I noticed this before I merged it (the CFLAGS issue, not the python thing). Even elsewhere in the build system things like -O3 and -g get added to CFLAGS, possibly overriding what the user has set. Ultimately it will be an upstream issue, though, and will probably take some time to (a) figure out all the places where CFLAGS get set and (b) convince upstream that they should only be set if no other flags are set by the user. The current ebuild is such an improvement over the others in the tree that it was worth it IMO.
Well, the QA is only about the python module so far. So, that particular bit will resolve itself in the next release. The rest of the mess doesn't trigger QA but is indeed very messy.
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=75e9a4b859d719f69fbda1138e87db624197f503 commit 75e9a4b859d719f69fbda1138e87db624197f503 Author: Michael Orlitzky <mjo@gentoo.org> AuthorDate: 2021-06-01 11:48:51 +0000 Commit: Michael Orlitzky <mjo@gentoo.org> CommitDate: 2021-06-01 11:49:16 +0000 sci-mathematics/singular: fix CFLAGS/CXXFLAGS support, mostly. This build system is a maze, but it turns out that we can fix most of the C(XX)FLAGS problems simply by passing --disable-optimizationflags to ./configure. Closes: https://bugs.gentoo.org/777792 Package-Manager: Portage-3.0.18, Repoman-3.0.2 Signed-off-by: Michael Orlitzky <mjo@gentoo.org> sci-mathematics/singular/singular-4.2.0_p3.ebuild | 1 + 1 file changed, 1 insertion(+)