RE: Run test on lsbsi-ppc64

"Wichmann, Mats D" <[email protected]> Thu, 1 Apr 2004 06:11:26 -0800
Newsgroups gmane.linux.lsb.implementation
Message-ID <[email protected]>
>For the powerpc64 platform, the lsbsi build can pass all the 4 phases
now.

Yes, I was able to do the same on build1-ppc64.linuxbase.org

>But it has some problems when run test_suite-2.0.2, lsblibchk with
lsbsi.

>1. Some symbols can not be found when run
lsb-libchk-1.9.0.20040328-1.ppc64.rpm
>   Many of the missing symbols is belong to libstdc++.so.5. The other
missed
>symbols are atexit (GLIBC_2.3) in libc.so.6 and pthread_atfork
(GLIBC_2.3) in
>libpthread.so.0

Don't worry about the libstdc++ symbols for now.  We're waiting for a
decision that affects this matter.

Just starting to explore the atexit issue.  We think, depending on
how the library is built, this may only appear in a static library
(libc_nonshared.a) for this glibc version.  The same is true for
pthread_atfork which may only be in the new libpthread_nonshared.a.
So don't worry about these either!  We've previously seen this on
x86_64 but not elsewhere (yet).

>2. Hang on /tset/PTHR.os/procprim/sigwait/T.sigwait
>This was also found in other platforms. Does anybody made some progress
on
>this problem?

This test fails everywhere, and hangs somewhere.  I haven't heard
an analysis of why it might hang - perhaps it's a mismatch between
library and kernel (I have an ia64 box where the test suite does
not hang on sigwait in native mode, but does hang in the lsbsi chroot).
It seems to hang on all Debian versions (native or lsbsi) so that
argues against the library mismatch with kernel theory though.

>3. Many failure and unresolved was found
>These problem need more investigation. Some of them are the same as
that
>on ia64-lsbsi

You have 86 failures which looks very similar to what I saw on ia64.
If you have a lot of unresolveds around syslog, you can make those
go away by stopping the native system's syslogd, and starting the
one inside the lsbsi chroot, before you start the test.  Those will
nearly all go away then.

>   Failure Breakdown:
>   Number of failures:     86      Number unresolved:      98
>   Number uninitiated:     4       Number unreported:      0