Re: gcc-3.4.X + glibc segfault

[Dan Kegel <dank at kegel dot com>]

Jake Page wrote:
> I found this message from a while back (like 2 months ago...):
>
> http://gcc.gnu.org/ml/gcc/2004-05/msg01148.html
>
> ...and was wondering if you ever figured out what was going on.

Yes, I think it was gcc-3.4.1 optimizing a bit much.
Here's a workaround backported from cvs to glibc-2.3.2:

http://kegel.com/crosstool/crosstool-0.28-rc26/patches/glibc-2.3.2/glibc-2.3.2-allow-gcc-3.4-nounit.patch

Hope that helps.

(I'm now trying to get glibc to build with gcc-3.5 snapshots,
which is a bit more interesting.  Lots of progress so far,
but currently stuck on inlining issues; gcc-3.5 doesn't like
it when you give both an inline and an out-of-line body for
a function declared with 'must inline' attributes...)
- Dan

> I am having the same problem with gcc-3.4.1/glibc-2.2.5 on MIPS.  I traced 
> it down to the exact same problem, in fact, these instructions:
>
>   mutex_init(&main_arena.mutex);
>   403838:       8f848020        lw      a0,-32736(gp)
>   40383c:       00000000        nop
>   403840:       24840904        addiu   a0,a0,2308
>   403844:       00002821        move    a1,zero
>   403848:       8f998770        lw      t9,-30864(gp)
>   40384c:       00000000        nop
>   403850:       0320f809        jalr    t9   <<== BOOM!
>
> $t9 = 0 at the jalr, so it jumps to 0x00000000, whoops!
>
> It looks like in thread-m.h mutex_init() is defined as:
>
> #define mutex_init(m)		\
>    (__pthread_mutex_init != NULL \
>     ? __pthread_mutex_init (m, NULL) : (*(int *)(m) = 0))
>
> where __pthread_mutex_init is a weak reference in libc and defined in 
> libpthread.  I also have a problem with this compiling shared 
> libs... however if I -lpthread shared it works, but static it still seg 
> faults.  Basically it looks like gcc-3.4.1 is "optimizing"?!? out the 
> conditional branch (though I tried -O0 and it still does it...)
>
> Anyway, probably more than you wanted to know :)  Just curious if you 
> found any more on it, if it's a gcc but I'm not too excited to try and fix 
> it, ugh...

--
My technical stuff: http://kegel.com
My politics: see http://www.misleader.org for examples of why I'm for regime change

------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com