Re: WARNING: program timed out.

[Segher Boessenkool <segher@kernel.crashing.org>]

Hi!

On Wed, May 06, 2020 at 11:55:58AM +0000, Dennis Clarke via Gcc-help wrote:
> I have a four stage bootstrap that took a very very very long time to
> complete.  Mostly due to the fact that I enabled code checks :
> 
> 
> --enable-bootstrap --enable-stage1-languages=c,c++ \
> --enable-checking=assert,misc,tree,gc,rtlflag,runtime,df,rtl,gcac,extra \
> --enable-stage1-checking=assert,misc,tree,gc,rtlflag,runtime,df,rtl,gcac,extra 
> \
> 
> 
> So this was fun to let it run and it did run where time -p reports :
> 
> .
> .
> .
> sparc64-sun-solaris2.10/sparcv8plus/libstdc++-v3/src/filesystem/cow-dir.o 
> differs
> sparc64-sun-solaris2.10/sparcv8plus/libstdc++-v3/src/filesystem/ops.o 
> differs
> sparc64-sun-solaris2.10/sparcv8plus/libstdc++-v3/src/filesystem/dir.o 
> differs
> sparc64-sun-solaris2.10/sparcv8plus/libstdc++-v3/src/filesystem/path.o 
> differs
> sparc64-sun-solaris2.10/sparcv8plus/libstdc++-v3/src/filesystem/cow-ops.o 
> differs
> sparc64-sun-solaris2.10/sparcv8plus/libstdc++-v3/src/c++17/fs_ops.o differs
> sparc64-sun-solaris2.10/sparcv8plus/libstdc++-v3/src/c++17/fs_dir.o differs
> sparc64-sun-solaris2.10/sparcv8plus/libstdc++-v3/src/c++17/cow-fs_path.o 
> differs
> sparc64-sun-solaris2.10/sparcv8plus/libstdc++-v3/src/c++17/cow-fs_dir.o 
> differs
> sparc64-sun-solaris2.10/sparcv8plus/libstdc++-v3/src/c++17/cow-fs_ops.o 
> differs
> sparc64-sun-solaris2.10/sparcv8plus/libstdc++-v3/src/c++17/fs_path.o differs
> gmake[2]: *** [Makefile:27282: compare3] Error 1
> gmake[2]: Leaving directory 
> '/opt/bw/build/gcc-9.2.0_SunOS5.10_sparc64vii+.003'
> gmake[1]: *** [Makefile:27262: stage4-bubble] Error 2
> gmake[1]: Leaving directory 
> '/opt/bw/build/gcc-9.2.0_SunOS5.10_sparc64vii+.003'
> gmake: *** [Makefile:27325: bootstrap4] Error 2
> 
> real 14078843.31
> user 13593892.79
> sys 437633.30
> beta$
> 
> I was surprised to see stage 3 and stage 4 differ in any way given that
>  stage 2 and stage 3 seemed perfect :
> 
> 
> beta$
> beta$ ls -ltrEin */xgcc */xg++
>    4387682 -rwxr-xr-x   1 16411    20002    8106984 2019-11-16 
> 22:53:28.194322200 +0000 stage1-gcc/xgcc
>    4388403 -rwxr-xr-x   1 16411    20002    8115680 2019-11-16 
> 22:53:50.102071100 +0000 stage1-gcc/xg++
>    4723452 -rwxr-xr-x   1 16411    20002    10271968 2019-12-30 
> 20:32:32.671996900 +0000 stage2-gcc/xgcc
>    4724042 -rwxr-xr-x   1 16411    20002    10284208 2019-12-30 
> 20:38:02.458504400 +0000 stage2-gcc/xg++
>    4871655 -rwxr-xr-x   1 16411    20002    10271968 2020-02-22 
> 08:38:28.492670800 +0000 prev-gcc/xgcc
>    4871806 -rwxr-xr-x   1 16411    20002    10284208 2020-02-22 
> 08:46:59.188336600 +0000 prev-gcc/xg++
>    4455418 -rwxr-xr-x   1 16411    20002    10271952 2020-04-16 
> 04:10:05.877113800 +0000 gcc/xgcc
>    4455567 -rwxr-xr-x   1 16411    20002    10284184 2020-04-16 
> 04:18:39.998805900 +0000 gcc/xg++
> beta$

Those are the compiler drivers, tiny programs compared to the compilers
themselves, which are called cc1, cc1plus, etc.

> May as well let the testsuite run ... however I see a ton of :
> 
> .
> .
> .
> FAIL: gcc.c-torture/execute/20040709-1.c   -O3 -fomit-frame-pointer 
> -funroll-loops -fpeel-loops -ftracer -finline-functions  (test for 
> excess errors)
> WARNING: program timed out.
> FAIL: gcc.c-torture/execute/20040709-1.c   -O3 -g  (test for excess errors)
> WARNING: program timed out.
> FAIL: gcc.c-torture/execute/20040709-1.c   -Os  (test for excess errors)
> WARNING: program timed out.
> FAIL: gcc.c-torture/execute/20040709-1.c   -O2 -flto 
> -flto-partition=none  (test for excess errors)
> WARNING: program timed out.
> .
> .
> .
> 
> I am guessing that there must be a "magic" testsuite incantation that
> allows the tests to run as slowly as they wish?  Anyone know?

1) Get a faster machine;
2) Change the timeouts (variables "timeout" and/or "gcc,timeout", in a
site.exp file perhaps);
3) Consider that the code might actually be in an infinite loop, the
default timeouts are quite generous, forever takes longer than any finite
stretch of time :-)


Segher