This is the mail archive of the
mailing list for the glibc project.
"make xcheck" questions
- From: Chris Metcalf <cmetcalf at tilera dot com>
- To: GLIBC Devel <libc-alpha at sourceware dot org>
- Date: Thu, 17 May 2012 14:57:11 -0400
- Subject: "make xcheck" questions
I'm working through the tilegx/tilepro tests and making sure I understand
any failure modes. A few things still confuse me:
1. Some of the "elf" tests try to build tst-unique3lib.so,
tst-unique3lib2.so, and tst-unique4lib.so. These fail to build on tile and
on x86_64, since they are compiled with g++ but then linked with gcc, so
there is a missing definition at link time of __gxx_personality_v0. I'm
configuring on x86_64 with: CFLAGS="-fasynchronous-unwind-tables -DNDEBUG
-g -O3" ~/git/glibc/configure --prefix=/usr --enable-add-ons=nptl,libidn
--disable-profile --enable-experimental-malloc. There may be some missing
magic here somewhere, but it does seem like if you compile with g++, you
need to link with g++.
2. I'm not seeing the c++-types-check.out test actually get run as part of
either "make xcheck" or "make tests". I can run it by hand (via "make
`pwd`/c++-types-check.out") and it does what I'd expect. I see that
"tests:" depends on this target so I'm confused as to how it's possible
that it's not running.
3. It would be nice if it were possible to explain to the c++-types-check
test that tilegx64 and tilegx32 should have different "data" files to check
against, but since both tilegx types have config-machine = tilegx, they end
up using the same "data" file. So the test passes on tilegx64 (the
"native" mode) and fails on tilegx32 (the "compat" 32-bit mode). Not
ideal, but certainly not a show-stopper.
Chris Metcalf, Tilera Corp.