make check problems on multilib systems in libstdc++

Benjamin Kosnik bkoz@redhat.com
Tue Jan 4 23:31:00 GMT 2005


Hey Andreas.



>The problem is that in libstdc++ a "make check" builds and installs
>the po files during "make check" in the default directory and then
>runs a "make all" in the other multilib directory ("32").  But during
>this "make all", this is not done:
>
>$ ls 32/libstdc++-v3/po/
>de.mo  fr.mo  Makefile
>
>$ ls libstdc++-v3/po/
>de.mo  fr.mo  Makefile  share
>
>So, what's the best way to do work around this?  We could install the
>compiled po files locally even with "make all".  Or is there another
>solution?

Ugh. I think the proper solution is to have the "make check" bits of
po/Makefile.am be multilib aware. In particular, this bit:

locale_builddir = $(glibcxx_localedir)

Please try the attached patch and see if it works.

>I noticed that on Linux/AMD64 which uses 32-bit x86 and 64-bit x86-64
>multilibs, some 32-bit tests fail.

>Btw. these missing compiled po files result in the following failures
>for 32-bit compilation (and using RUNTESTFLAGS="--target_board
>'unix{-m32,}'" for testing both multilibs):
>
>FAIL: 22_locale/messages/members/char/1.cc execution test
>FAIL: 22_locale/messages/members/char/2.cc execution test
>FAIL: 22_locale/messages/members/char/2.cc execution test
>FAIL: 22_locale/messages/members/char/wrapped_env.cc execution test
>FAIL: 22_locale/messages/members/char/wrapped_locale.cc execution test
>FAIL: 22_locale/messages_byname/named_equivalence.cc execution test

Perhaps this would also explain the ppc64 fails that Janis sees here:
http://gcc.gnu.org/ml/gcc-testresults/2005-01/msg00176.html

best,
benjamin
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: p.20050104
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20050104/cdc26b55/attachment.ksh>


More information about the Libstdc++ mailing list