Re: Results for egcs-2.92.22 19981122 (gcc2 ss-980609 experimental) testsuite on i586-pc-linux-gnu

[Nix <nix at esperi dot demon dot co dot uk>]

Jeffrey A Law <law@cygnus.com> writes:

> I bootstrapped with -O9 and -O1 again the other afternoon.
> 
> Precisely what problems are you having

-O9 ICEs in stage1 with

./xgcc -B./ -B/usr/packages.bin/egcs/snapshot/i586-pc-linux-gnu/bin/ -O2   -DIN_GCC    -W -Wall -W -Wall -O9 -funroll-all-loops -g -march=pentium -mcpu=pentium -fomit-frame-pointer -I./include  -fPIC -g1  -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED   -W -Wall -O9 -march=pentium -mcpu=pentium -fomit-frame-pointer -g -D__NO_MATH_INLINES -I. -I../../egcs/gcc -I../../egcs/gcc/config -I../../egcs/gcc/../include \
  -c -fexceptions ../../egcs/gcc/cp/exception.cc
../../egcs/gcc/cp/exception.cc: In function `void __check_eh_spec(int, const void **)':
../../egcs/gcc/cp/exception.cc:300: Internal compiler error.
../../egcs/gcc/cp/exception.cc:300: Please submit a full bug report to `egcs-bugs@cygnus.com'.

You suggested earlier that this might be a loop unrolling bug. My
testsuite scripts are totally shagged at the moment, though, so I
can't really tell ;) when they come back online (later today or
tomorrow, I hope) I'll see if this recurs if I turn loop unrolling
off.

-O1 segfaults in stage1:

_fixunssfdi
_fixsfdi
_fixxfdi
_fixunsxfdi
_floatdixf
_fixunsxfsi
_fixtfdi
_fixunstfdi
_floatditf
__gcc_bcmp
_varargs
__dummy
_eprintf
_bb
xgcc: Internal compiler error: program cc1 got fatal signal 11

I no longer have a command-line that causes one of these on record,
but when my testsuite scripts are unbroken I'll grab one of
them. Assuming these failures are still here, which wouldn't surprise
me (since I'm the only one they're biting).


>From my testsuite results, my environment and patches are

Patches:         - my patch of 981115 to libg++/etc/lf/entry.cc
Environment:     - GNU binutils 2.9.1.0.15
                 - dejagnu-980528
                 - gperf-2.7 with gperf-2.7-19981006 patch for -F
                 - Linux stable kernel 2.0.35 (compiled by gcc-2.7.2.3)
                 - glibc 2.0.6 (compiled by egcs-1.0.2)

>                                        and precisely how are you configuring
> and building

Scripted build with a BOOT_CFLAGS (broken to fit within 75cols) of
'-W -Wall -funroll-all-loops -g -march=pentium -mcpu=pentium \
 -fomit-frame-pointer -O[19]'

and a CFLAGS and CXXFLAGS of
'-W -Wall -march=pentium -mcpu=pentium -fomit-frame-pointer -g -O[19]'

Using bootstrap-lean.

> This will likely continue to fail until someone else can reproduce your
> problem.

I think you're right there.

-- 
`Anyone who says you can have a lot of widely dispersed people hack
 away on a complicated piece of code and avoid total anarchy has never
 managed a software project.' - Andy Tanenbaum in 1992 on comp.os.minix