Re: Debian Sarge and LSB V2.0 lsb-runtime test results

On Wed, 2005-06-15 at 15:33 -0400, Dave Neil wrote:
> Jeff Licquia wrote:
> >Would it be possible to post the journal?  That file should be in a
> >directory like "/home/tet/test_sets/results/0001e/journal".
> >
> No problem here you go;-)

My interpretation of the results:

The LI18NUX test failures are because of problems with wide character
handling in the upstream versions of diffutils and coreutils; since
upstream has not (to my knowledge) blessed a patch to fix them, Debian
has refused to incorporate the LSB patches.  In the past, the LSB has
waived those failures, and I had heard that the waiver would be granted
in the LSB 2.0 timeframe as well.

This leaves the following tests as problems:

> /tset/ANSI.os/time/ctime_X/T.ctime_X 1 FAIL
> /tset/ANSI.os/time/localtim_X/T.localtim_X 1 FAIL
> /tset/ANSI.os/time/mktime_X/T.mktime_X 1 FAIL
> /tset/PTHR.os/procprim/pthread_sigmask/T.pthread_sigmask 8 FAIL
> /tset/PTHR.os/procprim/pthread_sigmask/T.pthread_sigmask 9 FAIL
> /tset/PTHR.os/procprim/sigwait/T.sigwait 2 FAIL
> /tset/PTHR.os/procprim/sigwait/T.sigwait 5 UNREPORTED
> /tset/PTHR.os/thread/pthread_create/T.pthread_create 4 UNRESOLVED
> /tset/PTHR.os/thread/pthread_detach/T.pthread_detach 4 FAIL
> /tset/PTHR.os/thread/pthread_join/T.pthread_join 5 FAIL

The PTHR.os problems are caused by a slightly older version of NPTL in
Debian's glibc.  Porting the new NPTL is a pretty big job, I'm told, and
could result in new bugs popping up because of apps that depend on the
current behavior.

The time failures are caused by a bug in glibc regarding the handling of
daylight savings time.  The patch that fixes the problem, according to
the LSB wiki, is:

--- glibc-2.3.2/time/tzfile.c    2003-02-02 04:46:36.000000000 +0800
+++ glibc-2.3.2.lsb/time/tzfile.c    2004-04-06 02:31:05.000000000 +0800
@@ -414,6 +414,12 @@ __tzfile_default (const char *std, const
      as specified by this transition.  */
       isdst = trans_type->isdst;
     }
+ 
+  /* Now that we adjusted the transitions to the requested offsets,
+     reset the rule_stdoff and rule_dstoff values appropriately.  They
+     are used elsewhere.  */
+  rule_stdoff = stdoff;
+  rule_dstoff = dstoff;
 
   /* Reset types 0 and 1 to describe the user's settings.  */
   types[0].idx = 0;