GCC build goes into endless loop

GCC build goes into endless loop

[Gerald Pfeifer]

On i386-unknown-freebsd4.10 I started to see the following part of the 
bootstrap go into an endless loop (that is, not terminating after more 
than a day and being killed via a system ulimit):

ASM_HIDDEN_OP='' \
GCC_FOR_TARGET=' ./xgcc -B./ -B/sw/gcc-current/i386-unknown-freebsd4.10/bin/ -isystem /sw/gcc-current/i386-unknown-freebsd4.10/include -isystem /sw/gcc-current/i386-unknown-freebsd4.10/sys-include -L/files/pfeifer/OBJ-0814-0208/gcc/../ld' \
mkinstalldirs='/bin/sh /sw/test/gcc/cvs/gcc/../mkinstalldirs' \
  /bin/sh mklibgcc > tmp-libgcc.mk
mv tmp-libgcc.mk libgcc.mk
./xgcc -B./ -B/sw/gcc-current/i386-unknown-freebsd4.10/bin/ -isystem /sw/gcc-current/i386-unknown-freebsd4.10/include -isystem /sw/gcc-current/i386-unknown-freebsd4.10/sys-include -L/files/pfeifer/OBJ-0814-0208/gcc/../ld -O2 -O2 -g -O2   -DIN_GCC    -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -Wold-style-definition  -isystem ./include  -I. -I. -I/sw/test/gcc/cvs/gcc -I/sw/test/gcc/cvs/gcc/. -I/sw/test/gcc/cvs/gcc/../include -I./../intl -I/sw/test/gcc/cvs/gcc/../libcpp/include   -g0 -finhibit-size-directive -fno-inline-functions -fno-exceptions -fno-zero-initialized-in-bss -fno-unit-at-a-time   \
  -c /sw/test/gcc/cvs/gcc/crtstuff.c -DCRT_BEGIN \
  -o crtbegin.o
xgcc: Internal error: Terminated (program cc1)
Please submit a full bug report.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
gmake[2]: *** [crtbegin.o] Error 1
gmake[2]: Leaving directory `/files/pfeifer/OBJ-0814-0208/gcc'
gmake[1]: *** [stage2_build] Error 2
gmake[1]: Leaving directory `/files/pfeifer/OBJ-0814-0208/gcc'
gmake: *** [bootstrap-lean] Error 2

Can someone else reproduce this on an i386 or some other target?

Gerald

Re: GCC build goes into endless loop

[Andrew Pinski]

On Aug 15, 2005, at 4:03 PM, Gerald Pfeifer wrote:

> Can someone else reproduce this on an i386 or some other target?

This should been already fixed by:
2005-08-15  Sebastian Pop  <pop@cri.ensmp.fr>
         PR 23386

It was caused by:
2005-08-13  Sebastian Pop  <pop@cri.ensmp.fr>
         PR tree-optimization/22236

What is happening here is we were miss-compiling a finite loop to be an
infinite loop.


Thanks,
Andrew Pinski

Re: GCC build goes into endless loop

[Diego Novillo]

On Mon, Aug 15, 2005 at 10:03:33PM +0200, Gerald Pfeifer wrote:

> xgcc: Internal error: Terminated (program cc1)
> Please submit a full bug report.
> See <URL:http://gcc.gnu.org/bugs.html> for instructions.
> gmake[2]: *** [crtbegin.o] Error 1
> gmake[2]: Leaving directory `/files/pfeifer/OBJ-0814-0208/gcc'
> gmake[1]: *** [stage2_build] Error 2
> gmake[1]: Leaving directory `/files/pfeifer/OBJ-0814-0208/gcc'
> gmake: *** [bootstrap-lean] Error 2
> 
> Can someone else reproduce this on an i386 or some other target?
> 
I got something similar on last night's build, but it seems to
have been fixed now.  How recent is your tree?

Re: GCC build goes into endless loop

[Marcel Moolenaar]

On Aug 15, 2005, at 1:03 PM, Gerald Pfeifer wrote:

> ASM_HIDDEN_OP='' \
> GCC_FOR_TARGET=' ./xgcc -B./ -B/sw/gcc-current/i386-unknown- 
> freebsd4.10/bin/ -isystem /sw/gcc-current/i386-unknown-freebsd4.10/ 
> include -isystem /sw/gcc-current/i386-unknown-freebsd4.10/sys- 
> include -L/files/pfeifer/OBJ-0814-0208/gcc/../ld' \
> mkinstalldirs='/bin/sh /sw/test/gcc/cvs/gcc/../mkinstalldirs' \
>   /bin/sh mklibgcc > tmp-libgcc.mk
> mv tmp-libgcc.mk libgcc.mk
> ./xgcc -B./ -B/sw/gcc-current/i386-unknown-freebsd4.10/bin/ - 
> isystem /sw/gcc-current/i386-unknown-freebsd4.10/include -isystem / 
> sw/gcc-current/i386-unknown-freebsd4.10/sys-include -L/files/ 
> pfeifer/OBJ-0814-0208/gcc/../ld -O2 -O2 -g -O2   -DIN_GCC    -W - 
> Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -Wold- 
> style-definition  -isystem ./include  -I. -I. -I/sw/test/gcc/cvs/ 
> gcc -I/sw/test/gcc/cvs/gcc/. -I/sw/test/gcc/cvs/gcc/../include - 
> I./../intl -I/sw/test/gcc/cvs/gcc/../libcpp/include   -g0 -finhibit- 
> size-directive -fno-inline-functions -fno-exceptions -fno-zero- 
> initialized-in-bss -fno-unit-at-a-time   \
>   -c /sw/test/gcc/cvs/gcc/crtstuff.c -DCRT_BEGIN \
>   -o crtbegin.o
> xgcc: Internal error: Terminated (program cc1)
> Please submit a full bug report.
> See <URL:http://gcc.gnu.org/bugs.html> for instructions.
> gmake[2]: *** [crtbegin.o] Error 1
> gmake[2]: Leaving directory `/files/pfeifer/OBJ-0814-0208/gcc'

Gerald,

I see the same on ia64-hp-hpux11.23, hppa1.1-hp-hpux11.11 and hppa64- 
hp-hpux11.11.
I haven't had time to analyze the problem, though.

FYI,

-- 
  Marcel Moolenaar                                marcel@cup.hp.com

Re: GCC build goes into endless loop

[Sebastian Pop]

Andrew Pinski wrote:
> 
> On Aug 15, 2005, at 4:03 PM, Gerald Pfeifer wrote:
> 
> >Can someone else reproduce this on an i386 or some other target?
> 
> This should been already fixed by:
> 2005-08-15  Sebastian Pop  <pop@cri.ensmp.fr>
>         PR 23386
> 
> It was caused by:
> 2005-08-13  Sebastian Pop  <pop@cri.ensmp.fr>
>         PR tree-optimization/22236
> 
> What is happening here is we were miss-compiling a finite loop to be an
> infinite loop.
> 

Because niter estimation catches more cases now, it has triggered a
latent bug in the computation of the estimation.  Sorry again, next
time I'll modify scev I'll check on i686 also.

Sebastian

Re: GCC build goes into endless loop

[Gerald Pfeifer]

On Mon, 15 Aug 2005, Andrew Pinski wrote:
> This should been already fixed by:
> 2005-08-15  Sebastian Pop  <pop@cri.ensmp.fr>
>         PR 23386
> [...]
> What is happening here is we were miss-compiling a finite loop to be an
> infinite loop.

Thanks for the pointer, Andrew.

On Mon, 15 Aug 2005, Diego Novillo wrote:
> I got something similar on last night's build, but it seems to
> have been fixed now.  How recent is your tree?

It looked as if it was more recent then Sebastian's fix, but I found
that due to some gcc.gnu.org or network hickup not everything was 
completely up-to-date; tonights automated build on the same machine
suceeded again.

On Mon, 15 Aug 2005, Marcel Moolenaar wrote:
> I see the same on ia64-hp-hpux11.23, hppa1.1-hp-hpux11.11 and hppa64-
> hp-hpux11.11.
> I haven't had time to analyze the problem, though.

Does it work for you as well now?

Gerald

Re: GCC build goes into endless loop

[Marcel Moolenaar]

On Aug 17, 2005, at 12:09 AM, Gerald Pfeifer wrote:

> On Mon, 15 Aug 2005, Marcel Moolenaar wrote:
>
>> I see the same on ia64-hp-hpux11.23, hppa1.1-hp-hpux11.11 and hppa64-
>> hp-hpux11.11.
>> I haven't had time to analyze the problem, though.
>
> Does it work for you as well now?

I still have the failure with hppa64-hp-hpux11.11 (natively). The  
other 2
targets build fine now.

-- 
  Marcel Moolenaar                                marcel@cup.hp.com

Re: GCC build goes into endless loop

[Andrew Pinski]

On Aug 17, 2005, at 11:01 AM, Marcel Moolenaar wrote:

>
> On Aug 17, 2005, at 12:09 AM, Gerald Pfeifer wrote:
>
>> On Mon, 15 Aug 2005, Marcel Moolenaar wrote:
>>
>>> I see the same on ia64-hp-hpux11.23, hppa1.1-hp-hpux11.11 and hppa64-
>>> hp-hpux11.11.
>>> I haven't had time to analyze the problem, though.
>>
>> Does it work for you as well now?
>
> I still have the failure with hppa64-hp-hpux11.11 (natively). The 
> other 2
> targets build fine now.

Works for CodeSourcery just fine:
http://gcc.gnu.org/ml/gcc-testresults/2005-08/msg00931.html

Are you sure that you are building with "make bootstrap".

-- Pinski

Re: GCC build goes into endless loop

[Marcel Moolenaar]

On Aug 17, 2005, at 8:05 AM, Andrew Pinski wrote:

>
> On Aug 17, 2005, at 11:01 AM, Marcel Moolenaar wrote:
>
>
>>
>> On Aug 17, 2005, at 12:09 AM, Gerald Pfeifer wrote:
>>
>>
>>> On Mon, 15 Aug 2005, Marcel Moolenaar wrote:
>>>
>>>
>>>> I see the same on ia64-hp-hpux11.23, hppa1.1-hp-hpux11.11 and  
>>>> hppa64-
>>>> hp-hpux11.11.
>>>> I haven't had time to analyze the problem, though.
>>>>
>>>
>>> Does it work for you as well now?
>>>
>>
>> I still have the failure with hppa64-hp-hpux11.11 (natively). The  
>> other 2
>> targets build fine now.
>>
>
> Works for CodeSourcery just fine:
> http://gcc.gnu.org/ml/gcc-testresults/2005-08/msg00931.html

Ok, thanks for the pointer. I'll double-check and try to spot
the differences. For now, it's probably best to assume it's
something local here.

> Are you sure that you are building with "make bootstrap".

Positive.

-- 
  Marcel Moolenaar                                marcel@cup.hp.com