I don't know the build system for this program (and this is a generic bug template so I cannot tell you which program exactly is), but my tests shows that it's not respecting CFLAGS (or CXXFLAGS) properly. Please look into it, since it's important to respect user CFLAGS and CXXFLAGS. Warning: this bug might look like a false positive because you actually have your CFLAGS being used; this happens if the CFLAGS are "set in stone" in the build system during src_unpack/src_prepare. While QA has not as of this moment expressed to me a preference, I'd sincerely suggest to avoid the set-in-stone approach, so that ebuild commands could work to reproduce the actual results. To avoid the set in stone approach: - consider just changing CFLAGS= to CFLAGS+= if the build system enables warnings; - if the buildsystem does not use CFLAGS variable at all, in the sed use '$(CFLAGS)', single quoted, so that the CFLAGS variable is picked up; - use '$(OPTCFLAGS)' in the sed and then use make OPTCFLAGS=$CFLAGS. Thanks, Diego
This is fixed in 9.4.3-r4, along with proper support for CC and LDFLAGS. During one file, *all* CFLAGS (even those set by the package) were deliberately omitted because even -O caused problems, according to the comment. I've updated this to use ${CFLAGS} -O0. Additionally, the need for -fwrapv is exposed at -O2, so I've added that as well. If you have better ideas for either of those, please do add them here.