New Xorg crashes when built with clang

[Volodymyr Kostyrko]

Gautam wrote:
> Hi,
>
> On Mon, May 21, 2012 at 3:40 PM, Gautam<list at execve.net>  wrote:
>
>> #0  0x0000000802bdc43c in thr_kill () from /lib/libc.so.7
>> #1  0x0000000802c7582c in abort () from /lib/libc.so.7
>> #2  0x000000000046d51e in OsAbort ()
>> #3  0x0000000000478925 in ddxGiveUp ()
>> #4  0x000000000046accd in AbortServer ()
>> #5  0x000000000046a4b2 in FatalError ()
>> #6  0x000000000046b6be in OsInit ()
>> #7  0x0000000802971b68 in swapcontext () from /lib/libthr.so.3
>> #8  0x000000080297170f in sigaction () from /lib/libthr.so.3
>> #9  0x00007ffffffff043 in ?? ()
>> #10 0x00000008029715e0 in sigaction () from /lib/libthr.so.3
>> #11 0x00000008040a1d54 in ProcDRI2Dispatch ()
>>     from /usr/local/lib/xorg/modules/extensions/libdri2.so
>> #12 0x0000000000433bdd in Dispatch ()
>> #13 0x000000000042dbf4 in main ()
>> (gdb)

Try updating world. clang 3.1 is in the tree.

This might be just an itchy feeling, but clang 3.0 may produce very 
weird code when compiling with -march=native. I faced two weirdness samples:

  * on i386 with atlon-xp processor world built with -march=native fails 
to load libraries dynamically and rolls clock back into the 1942.

  * on some intel chips amd64 world built with -march=native was 
unusable by mysql (but it seems 0k with some other chips). I faced that 
on E5520. And the error was in libthr.so.3 too.

If you have built your world with -march=native you can succumb to one 
of those. However latest clang import seems to fix at least first one 
for me.

-- 
Sphinx of black quartz judge my vow.