glibc 2.1.93

[Franz Sirl]

At 14:39 20.09.00, Andreas Jaeger wrote:
> >>>>> Christian Iseli writes:
>
>  > Hi folks,
>  > geoffk@cygnus.com said:
> >> This is OK, because the output from the compiler is fed directly into
> >> the assembler; there's no preprocessor in the way.
>
>  > Ok, things look brighter today.  With current GCC CVS (20000920), and 
> current
>  > libc CVS, I have:
>
>  > $ make -k check >& check.log
>  > $ grep '\*\*' check.log
>  > make[2]: *** 
> [/usr/src/redhat/BUILD/glibc-build-i386/localedata/tst-locale.out]
>  >  Error 139
>  > make[1]: *** [localedata/tests] Error 2
>  > make[2]: *** 
> [/usr/src/redhat/BUILD/glibc-build-i386/debug/backtrace-tst.out]
>  >  Error 139
>  > make[1]: *** [debug/tests] Error 2
>  > make: *** [check] Error 2
>
>Looks like a compiler error but I'm not sure.

Yes, I think so too. On powerpc-linux-gnu all is fine if I compile with my 
latest gcc-2.95.3-2d (includes lots of mainline patches backported), but I 
get the same localedata segfault with yesterdays mainline gcc. I think this 
is a fallout of the recent introduction of the 
function-at-a-time-processing in the mainline.

Franz. (who is now heading to the official company meeting at the 
Oktoberfest :-) )