Re: gnustep-base tests

[Eric Heintzmann]

Le 11/05/2016 08:58, Richard Frith-Macdonald a écrit :
> > On 10 May 2016, at 23:15, Eric Heintzmann <heintzmann.eric@free.fr> wrote:
> >
> >
> >
> > Le 10/05/2016 11:41, Richard Frith-Macdonald a écrit :
> > > On the one hand, the particular number format stuff is rarely used, so a bug 
> > > there may not be important (and may depend on the particular ICU library 
> > > shipped on the system).
> > Well I ve done some tests and this particular problem is common to debian 
> > testing (stretch)  and unstable (sid) and Ubuntu 16.04 too...
> > > On the other, we really want to fix *any* failure of the regression tests.
> > > Does the same problem occur with gnustep-base from svn trunk?
> > > Can you get a stack trace of the crash?
> > Well I ve tried to use gnustep-make 2.6.6-3 to build gnustep-base 1.24.9 and 
> > the regrssesion test ar ok !!
> > It seem the bug is in gnustep-make 2.6.8 (or in the packaging)
> While I don't remember anything significant changng, I guess there might have 
> been a change in the test framework code which could have exposed a bug in the 
> library  not noticed in the older version or something.  I think we need stack 
> traces from gdb (ie full symbolic information) to find out where the issue is.  
> It's also strange that nobody else is seeing it.
> I'm going to try installing a new debian system to see if I can reproduce it, 
> but that's going to take quite some time, and I don't think the chances are 
> very good (too many unknown variables).
I've started to play with gdb.
After the crash there are these strange lines:

Program received signal SIGABRT, Aborted.
0x00007ffff6925478 in __GI_raise (sig=sig@entry=6)
    at ../sysdeps/unix/sysv/linux/raise.c:55
55      ../sysdeps/unix/sysv/linux/raise.c: No such file or directory.