Parallel build problems on sparc-sun-solaris2.8

Parallel build problems on sparc-sun-solaris2.8

[Gerald Pfeifer]

Has any of you access to a two-processor SPARC box running Solaris 2.8
(or perhaps 2.7)?

I'm getting weird parallel build failures and would like to know whether
these can be reproduced by someone else using my settings.

(Perhaps this even can be reproduced on a different two-processor box.
If someone would like to give it a try, please contact me directly.)

Gerald
-- 
Gerald "Jerry" pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/

Re: Parallel build problems on sparc-sun-solaris2.8

[H . J . Lu]

On Wed, Sep 20, 2000 at 03:12:48PM +0200, Gerald Pfeifer wrote:
> Has any of you access to a two-processor SPARC box running Solaris 2.8
> (or perhaps 2.7)?
> 
> I'm getting weird parallel build failures and would like to know whether
> these can be reproduced by someone else using my settings.
> 
> (Perhaps this even can be reproduced on a different two-processor box.
> If someone would like to give it a try, please contact me directly.)
> 

Parallel build has been broken for a while. I reported it and provided
a kludge. But very few people seem to care.



H.J.

Re: Parallel build problems on sparc-sun-solaris2.8

[Gerald Pfeifer]

On Sat, 23 Sep 2000, Phil Edwards wrote:
> Then I remembered that I had forgotten to upgrade my gmake.  Once I had
> done that, using gmake 3.79.1, the exact same previously-successful source
> fails a fresh bootstrap with

It seems that this problem has been fixed (or ``disappeared'') now, at
least that's what several tests this weekend have shown.

Thanks to you and others who helped with verifying this problem!

Gerald
-- 
Gerald "Jerry" pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/

Re: Parallel build problems on sparc-sun-solaris2.8

[Gerald Pfeifer]

On Sat, 23 Sep 2000, H . J . Lu wrote:
>> Would you mind reposting that patch? I'll definitely give it a try.
> http://gcc.gnu.org/ml/gcc-bugs/2000-08/msg00557.html

Thanks! I tried this, but it didn't solve the problem we (i.e., myself
and others) have been seing on sparc-sun-solaris2.8.

I'm investigating...

Gerald
-- 
Gerald "Jerry" pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/

Re: Parallel build problems on sparc-sun-solaris2.8

[Phil Edwards]

On Sat, Sep 23, 2000 at 11:29:57AM +0200, Gerald Pfeifer wrote:
> but I'm also getting 
> 
>   gmake[4]: *** No rule to make target `genemit.c', needed by `genemit.o'.  Stop.
>   gmake[4]: *** Waiting for unfinished jobs....
>   gmake[4]: Leaving directory `/files/pfeifer/OBJ-2009-23:12/gcc'
>   gmake[3]: *** [sparcv9/gmon.o] Error 2
>   gmake[3]: *** Waiting for unfinished jobs....
>   ./xgcc -B./ -B/sw/test/gcc/SunOS/sparc-sun-solaris2.8/bin/ -isystem 
> 
> and also different ones. At least one other tester has now been able to
> reproduce this problem.

As the tester in question, I'll post my results for the record.  A quick
reminder to all, though, that recursive parallel make handling changed in
gmake 3.79.

Using Solaris 8.1 on a dual-CPU box with gmake 3.77 and -j4, I was unable
to reproduce the problem; current CVS built fine.

Then I remembered that I had forgotten to upgrade my gmake.  Once I had
done that, using gmake 3.79.1, the exact same previously-successful source
fails a fresh bootstrap with

   No rule to make target `lists.c', needed by `lists.o'.  Stop.

The last several lines of the build are included below.

Phil

-- 
pedwards at disaster dot jaj dot com  |  pme at sources dot redhat dot com
devphil at several other less interesting addresses in various dot domains
The gods do not protect fools.  Fools are protected by more capable fools.


gmake[4]: Leaving directory `/tmp/pedwards/gerald/gcc'
mv opnew.o libgcc/sparcv9/opnew.o
gmake GCC_FOR_TARGET="./xgcc -B./ -B/home/pedwards/geraldtest/sparc-sun-solaris2.8/bin/ -isystem /home/pedwards/geraldtest/sparc-sun-solaris2.8/include" \
  AR_FOR_TARGET="ar" \
  AR_CREATE_FOR_TARGET="ar  rc" \
  AR_EXTRACT_FOR_TARGET="" \
  AR_FLAGS_FOR_TARGET="" \
  CC="stage1/xgcc -Bstage1/ -B/home/pedwards/geraldtest/sparc-sun-solaris2.8/bin/" CFLAGS="-O2 -g -O2 -W -Wall -Wtraditional -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -pedantic -Wno-long-long" \
  HOST_PREFIX="" \
  HOST_PREFIX_1="loser-" \
  LANGUAGES="c gcov c++ java" \
  LIBGCC2_CFLAGS="-O2   -DIN_GCC  -DSVR4  -O2 -g -O2 -W -Wall  -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes  -Wno-long-long -isystem ./include  -fPIC -g1 -DHAVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  -m64 "  \
  MULTILIB_CFLAGS=" -m64" opvnewnt.o
gmake[4]: Leaving directory `/tmp/pedwards/gerald/gcc'
mv opnewnt.o libgcc/sparcv9/opnewnt.o
gmake GCC_FOR_TARGET="./xgcc -B./ -B/home/pedwards/geraldtest/sparc-sun-solaris2.8/bin/ -isystem /home/pedwards/geraldtest/sparc-sun-solaris2.8/include" \
  AR_FOR_TARGET="ar" \
  AR_CREATE_FOR_TARGET="ar  rc" \
  AR_EXTRACT_FOR_TARGET="" \
  AR_FLAGS_FOR_TARGET="" \
  CC="stage1/xgcc -Bstage1/ -B/home/pedwards/geraldtest/sparc-sun-solaris2.8/bin/" CFLAGS="-O2 -g -O2 -W -Wall -Wtraditional -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -pedantic -Wno-long-long" \
  HOST_PREFIX="" \
  HOST_PREFIX_1="loser-" \
  LANGUAGES="c gcov c++ java" \
  LIBGCC2_CFLAGS="-O2   -DIN_GCC  -DSVR4  -O2 -g -O2 -W -Wall  -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes  -Wno-long-long -isystem ./include  -fPIC -g1 -DHAVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  -m64 "  \
  MULTILIB_CFLAGS=" -m64" opdel.o
gmake[4]: Entering directory `/tmp/pedwards/gerald/gcc'
gmake[4]: Entering directory `/tmp/pedwards/gerald/gcc'
gmake[4]: Leaving directory `/tmp/pedwards/gerald/gcc'
mv tinfo2.o libgcc/sparcv9/tinfo2.o
gmake GCC_FOR_TARGET="./xgcc -B./ -B/home/pedwards/geraldtest/sparc-sun-solaris2.8/bin/ -isystem /home/pedwards/geraldtest/sparc-sun-solaris2.8/include" \
  AR_FOR_TARGET="ar" \
  AR_CREATE_FOR_TARGET="ar  rc" \
  AR_EXTRACT_FOR_TARGET="" \
  AR_FLAGS_FOR_TARGET="" \
  CC="stage1/xgcc -Bstage1/ -B/home/pedwards/geraldtest/sparc-sun-solaris2.8/bin/" CFLAGS="-O2 -g -O2 -W -Wall -Wtraditional -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -pedantic -Wno-long-long" \
  HOST_PREFIX="" \
  HOST_PREFIX_1="loser-" \
  LANGUAGES="c gcov c++ java" \
  LIBGCC2_CFLAGS="-O2   -DIN_GCC  -DSVR4  -O2 -g -O2 -W -Wall  -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes  -Wno-long-long -isystem ./include  -fPIC -g1 -DHAVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  -m64 "  \
  MULTILIB_CFLAGS=" -m64" opdelnt.o
gmake[4]: *** No rule to make target `lists.c', needed by `lists.o'.  Stop.
gmake[4]: *** Waiting for unfinished jobs....
gmake[4]: Leaving directory `/tmp/pedwards/gerald/gcc'
gmake[3]: *** [libgcc/sparcv9/opvnewnt.o] Error 2
gmake[3]: *** Waiting for unfinished jobs....
gmake[4]: Leaving directory `/tmp/pedwards/gerald/gcc'
mv opvnew.o libgcc/sparcv9/opvnew.o
gmake[3]: *** Waiting for unfinished jobs....
gmake[4]: Entering directory `/tmp/pedwards/gerald/gcc'
./xgcc -B./ -B/home/pedwards/geraldtest/sparc-sun-solaris2.8/bin/ -isystem /home/pedwards/geraldtest/sparc-sun-solaris2.8/include -O2   -DIN_GCC  -DSVR4  -O2 -g -O2 -W -Wall  -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes  -Wno-long-long -isystem ./include  -fPIC -g1 -DHAVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  -m64   -g -O2 -I. -I/home/pedwards/src/egcs/gcc -I/home/pedwards/src/egcs/gcc/config -I/home/pedwards/src/egcs/gcc/../include \
  -c /home/pedwards/src/egcs/gcc/cp/new2.cc -DL_op_delete -o opdel.o
./xgcc -B./ -B/home/pedwards/geraldtest/sparc-sun-solaris2.8/bin/ -isystem /home/pedwards/geraldtest/sparc-sun-solaris2.8/include -O2   -DIN_GCC  -DSVR4  -O2 -g -O2 -W -Wall  -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes  -Wno-long-long -isystem ./include  -fPIC -g1 -DHAVE_GTHR_DEFAULT -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  -m64   -g -O2 -I. -I/home/pedwards/src/egcs/gcc -I/home/pedwards/src/egcs/gcc/config -I/home/pedwards/src/egcs/gcc/../include \
  -c /home/pedwards/src/egcs/gcc/cp/new2.cc -DL_op_delnt -o opdelnt.o
gmake[4]: Leaving directory `/tmp/pedwards/gerald/gcc'
mv opdel.o libgcc/sparcv9/opdel.o
gmake[4]: Leaving directory `/tmp/pedwards/gerald/gcc'
mv opdelnt.o libgcc/sparcv9/opdelnt.o
gmake[3]: Leaving directory `/tmp/pedwards/gerald/gcc'
gmake[2]: *** [stmp-multilib] Error 2
gmake[2]: Leaving directory `/tmp/pedwards/gerald/gcc'
gmake[1]: *** [stage_c] Error 2
gmake[1]: Leaving directory `/tmp/pedwards/gerald/gcc'
gmake: *** [bootstrap] Error 2

Re: Parallel build problems on sparc-sun-solaris2.8

[H . J . Lu]

On Sat, Sep 23, 2000 at 11:29:57AM +0200, Gerald Pfeifer wrote:
> On Wed, 20 Sep 2000, H . J . Lu wrote:
> > Parallel build has been broken for a while. I reported it and provided
> > a kludge. But very few people seem to care.
> 
> Would you mind reposting that patch? I'll definitely give it a try.

http://gcc.gnu.org/ml/gcc-bugs/2000-08/msg00557.html

It seems that it is expected to patch gcc for parallel build on smp.

BTW, I haven't bootstraped gcc for quite a few days. I don't know
what else is broken now :-).


H.J.

Re: Parallel build problems on sparc-sun-solaris2.8

[Gerald Pfeifer]

On Wed, 20 Sep 2000, H . J . Lu wrote:
> Parallel build has been broken for a while. I reported it and provided
> a kludge. But very few people seem to care.

Would you mind reposting that patch? I'll definitely give it a try.

On Wed, 20 Sep 2000, Mike Stump wrote:
> Most parallel build problems can be seen on single processor boxes
> with a -j3 to -j5 bootstrap. [...] They can usually be identified by
> the fact that make -j1 works flawlessly, and a make -j5 doesn't.  :-(

make -j1 on the dual-processor box works fine, and also make -j4 on a
similiar single-processor box does (both tested half a dozen times).

My first report is at http://gcc.gnu.org/ml/gcc-bugs/2000-09/msg00354.html ,
but I'm also getting 

  gmake[4]: *** No rule to make target `genemit.c', needed by `genemit.o'.  Stop.
  gmake[4]: *** Waiting for unfinished jobs....
  gmake[4]: Leaving directory `/files/pfeifer/OBJ-2009-23:12/gcc'
  gmake[3]: *** [sparcv9/gmon.o] Error 2
  gmake[3]: *** Waiting for unfinished jobs....
  ./xgcc -B./ -B/sw/test/gcc/SunOS/sparc-sun-solaris2.8/bin/ -isystem 

and also different ones. At least one other tester has now been able to
reproduce this problem.


The only difference between my two boxes is the number of processors
and the version of Solaris: 2.8 vs 2.6, so this is either a bug in our
Makefiles(?), a bug in GNU make or in Solaris. :-(

Gerald
-- 
Gerald "Jerry" pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/

Re: Parallel build problems on sparc-sun-solaris2.8

[Mike Stump]

> Date: Wed, 20 Sep 2000 15:12:48 +0200 (CEST)
> From: Gerald Pfeifer <pfeifer@dbai.tuwien.ac.at>
> To: gcc@gcc.gnu.org

> Has any of you access to a two-processor SPARC box running Solaris
> 2.8 (or perhaps 2.7)?

> I'm getting weird parallel build failures and would like to know whether
> these can be reproduced by someone else using my settings.

Most parallel build problems can be seen on single processor boxes
with a -j3 to -j5 bootstrap.  They can usually be identified by the
fact that make -j1 works flawlessly, and a make -j5 doesn't.  :-(

If you show us the error in a make -j5, we can probably tell you if
we've seen it before.