Re: Mainline build problem on FC4

[Andrew MacLeod <amacleod at redhat dot com>]

On Mon, 2006-06-19 at 12:31 -0400, Andrew MacLeod wrote: 
> OK, so I wasn't just wacko last week.  Mainline fails to build on Fedora
> Core 4 (with all the latest updates,  kernel 2111) when built with
> checking disabled.  Diego verified this on a second FC4 box.  
> 

I've narrowed this down to the patch in revision 114674.  (114673 works
fine, 114674 exhibits the problem) 

------------------------------------------------------------------------
r114674 | rakdver | 2006-06-15 05:42:03 -0400 (Thu, 15 Jun 2006) | 10 lines

        * tree-ssa-loop-niter.c (implies_nonnegative_p): New function.
        (derive_constant_upper_bound): Derive more precise upper bound in
        common cases.  Return type changed to double_int.
        (record_estimate): Reflect the changed return type of
        derive_constant_upper_bound.
        * double-int.c (double_int_zext, double_int_sext): Fix.

        * gcc.dg/tree-ssa/loop-18.c: New test.



bitmap.o in stage 2 is miscompiled by the stage 1 compiler. If I replace
bitmap.o with the bitmap.o produced by the stage 2 compiler from rev.
114673, everything proceeds and we get a failure in stage3.

Thats as far as I've gotten today. 

Andrew

PS. The only configure options are --enable-languages=c,c++
--disable-checking

> The compile fails when building crtfastmath during stage2.
> 
> It doesn't happen when checking is enabled, nor does it happen on FC5
> under any circumstance I have found.
> 
> 
> /build/gcc/2006-06-19-nc/./gcc/xgcc -B/build/gcc/2006-06-19-nc/./gcc/ -B/install/gcc-nc/i686-pc-linux-gnu/bin/ -B/install/gcc-nc/i686-pc-linux-gnu/lib/ -isystem /install/gcc-nc/i686-pc-linux-gnu/include -isystem /install/gcc-nc/i686-pc-linux-gnu/sys-include -O2  -O2 -g -O2  -DIN_GCC    -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -Wold-style-definition  -isystem ./include  -fPIC -g -DHAVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  -msse -c \
>         /src/gcc/2006-06-19/gcc/gcc/config/i386/crtfastmath.c \
>         -o crtfastmath.o
> /bin/sh /src/gcc/2006-06-19/gcc/gcc/../move-if-change tmp-macro_list macro_list
> *** glibc detected *** /build/gcc/2006-06-19-nc/./gcc/cc1: malloc(): memory corruption: 0x08591630 ***
> ======= Backtrace: =========
> /lib/libc.so.6[0x5ee1a2]
> /lib/libc.so.6(malloc+0x74)[0x5ef587]
> /lib/libc.so.6[0x5af193]
> /lib/libc.so.6[0x5ad498]
> /lib/libc.so.6[0x5ace25]
> /lib/libc.so.6(__dcgettext+0x43)[0x5abed7]
> /lib/libc.so.6(strsignal+0x13c)[0x5f4a13]
> /build/gcc/2006-06-19-nc/./gcc/cc1[0x830e5b2]
> ======= Memory map: ========
> 0055d000-0055e000 r-xp 0055d000 00:00 0          [vdso]
> 0055e000-00578000 r-xp 00000000 fd:00 5499650    /lib/ld-2.3.6.so
> 00578000-00579000 r-xp 00019000 fd:00 5499650    /lib/ld-2.3.6.so
> 00579000-0057a000 rwxp 0001a000 fd:00 5499650    /lib/ld-2.3.6.so
> 0058a000-006ad000 r-xp 00000000 fd:00 5499657    /lib/libc-2.3.6.so
> 006ad000-006af000 r-xp 00122000 fd:00 5499657    /lib/libc-2.3.6.so
> 006af000-006b1000 rwxp 00124000 fd:00 5499657    /lib/libc-2.3.6.so
> 006b1000-006b3000 rwxp 006b1000 00:00 0
> 00a29000-00a33000 r-xp 00000000 fd:00 16893261   /build/gcc/2006-06-19-nc/prev-gcc/libgcc_s.so.1
> 00a33000-00a34000 rwxp 00009000 fd:00 16893261   /build/gcc/2006-06-19-nc/prev-gcc/libgcc_s.so.1
> 08048000-084cf000 r-xp 00000000 fd:00 16926839   /build/gcc/2006-06-19-nc/gcc/cc1
> 084cf000-084d4000 rw-p 00486000 fd:00 16926839   /build/gcc/2006-06-19-nc/gcc/cc1
> 084d4000-085a0000 rw-p 084d4000 00:00 0          [heap]
> b7b00000-b7b21000 rw-p b7b00000 00:00 0
> b7b21000-b7c00000 ---p b7b21000 00:00 0
> b7c87000-b7dd8000 rw-p b7c87000 00:00 0
> b7dd8000-b7fd8000 r--p 00000000 fd:00 201117     /usr/lib/locale/locale-archive
> b7fd8000-b7fd9000 rw-p b7fd8000 00:00 0
> b7ffe000-b8000000 rw-p b7ffe000 00:00 0
> bffe8000-bffff000 rw-p bffe8000 00:00 0          [stack]
> xgcc: Internal error: Segmentation fault (program cc1)
> Please submit a full bug report.
> See <URL:http://gcc.gnu.org/bugs.html> for instructions.
> make[3]: *** [crtfastmath.o] Error 1
> make[3]: *** Waiting for unfinished jobs....
> echo timestamp > s-macro_list
> make[3]: *** Waiting for unfinished jobs....
> rm fsf-funding.pod gcov.pod gfdl.pod cpp.pod gpl.pod gcc.pod
> make[3]: Leaving directory `/build/gcc/2006-06-19-nc/gcc'
> make[2]: *** [all-stage2-gcc] Error 2
> make[2]: Leaving directory `/build/gcc/2006-06-19-nc'
> make[1]: *** [stage2-bubble] Error 2
> make[1]: Leaving directory `/build/gcc/2006-06-19-nc'
> make: *** [all] Error 2
>  
> 
> 
> Has anyone else encountered this?  I'm checking back to see when/if this
> suddenly appeared.  It also happens with a much earlier version of the
> kernel. If I recall from my poking around earlier last week, bitmap.o
> get miscompiled in stage1, but take that with a grain of salt.. a lot of
> things were going on last week and I don't know if that is still true
> today.
> 
> Andrew
>