[Debian-med-packaging] Bug#917851: Failed build for seqan2 on i386

[Aaron M. Ucko]

Andreas Tille <tille at debian.org> writes:

> But other 32bit architectures like armel and armhf are passing[2].  So I
> fail to see why exactly i386 is failing.  Is this possibly an effect of
> bug #917851?

Probably not; dropping the bug to a Bcc.  Experimentation in an i386
chroot reveals the problem to be specifically with yara_mapper, which
https://salsa.debian.org/med-team/seqan2/-/blob/master/debian/patches/skip-some-apps-on-some-archs
explicitly excludes (along with yara_indexer) on several other 32-bit
platforms.  We could go the same route for i386, but AFAICT it suffices
to drop the optimization level back down to -O2 for this specific
application, by adding

# Drop back from global -O3 on i386 to avoid
# "virtual memory exhausted: Operation not permitted"
if ("$ENV{DEB_BUILD_ARCH}" STREQUAL "i386")
    target_compile_options (yara_mapper PRIVATE "-O2")
endif ()

to apps/yara/CMakeLists.txt following the add_executable call for
yara_mapper.  (If and when debian/rules honors noopt, we should further
conditionalize this call accordingly, but I'm not familiar enough with
cmake to come up with the correct syntax offhand.)  We could perhaps try
doing the same for other affected platforms in an upload to
experimental.

-- 
Aaron M. Ucko, KB1CJC (amu at alum.mit.edu, ucko at debian.org)
http://www.mit.edu/~amu/ | http://stuff.mit.edu/cgi/finger/?amu@monk.mit.edu