egcs-release branch aborts during bootstrap

egcs-release branch aborts during bootstrap

[Robert Lipe]

[ Manfred, does this fail on your COFF target? ] 

This command works on the trunk.   It fails on the release tag.
The sources of the release branch were checked out last night, and
a 'cvs update -n' in gcc/cp doesn't show any updates pending.  It
shows up in a make boostrap2 on OpenServer.

(robertl) rjlhome:/play/egcs-release/native/gcc
$ cat /tmp/cc
./xgcc -B./ -O2   -DIN_GCC    -O2 -g -I./include   -g1 \
        -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -fexceptions    \
        -I. -I../../egcs/gcc -I../../egcs/gcc/config -mcoff \
          -c ../../egcs/gcc/cp/tinfo.cc
(robertl) rjlhome:/play/egcs-release/native/gcc
$ sh /tmp/cc
include/exception: In method `bad_exception::bad_exception()':
In file included from include/typeinfo:7,
                 from ../../egcs/gcc/cp/tinfo.h:4,
                 from ../../egcs/gcc/cp/tinfo.cc:29:
include/exception:24: Internal compiler error.
include/exception:24: Please submit a full bug report to `egcs-bugs@cygnus.com'.

If I change '-fexceptions' to '-fno-exceptions' it doesn't crash.

If I remove -mcoff, (defaulting back to ELF) it doesn't crash.  The core
dumps aren't readable by GDB.  Whether this is a GDB problem or an EGCS
problem is a different mystery.

-- 
Robert Lipe       http://www.dgii.com/people/robertl       robertl@dgii.com

Re: egcs-release branch aborts during bootstrap

[Manfred.Hollstein]

On Thu, 12 February 1998, 15:28:22, robertl@dgii.com wrote:

 > [ Manfred, does this fail on your COFF target? ] 
 > 
 > This command works on the trunk.   It fails on the release tag.
 > The sources of the release branch were checked out last night, and
 > a 'cvs update -n' in gcc/cp doesn't show any updates pending.  It
 > shows up in a make boostrap2 on OpenServer.
 > 
 > (robertl) rjlhome:/play/egcs-release/native/gcc
 > $ cat /tmp/cc
 > ./xgcc -B./ -O2   -DIN_GCC    -O2 -g -I./include   -g1 \
 >         -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -fexceptions    \
 >         -I. -I../../egcs/gcc -I../../egcs/gcc/config -mcoff \
 >           -c ../../egcs/gcc/cp/tinfo.cc
 > (robertl) rjlhome:/play/egcs-release/native/gcc
 > $ sh /tmp/cc
 > include/exception: In method `bad_exception::bad_exception()':
 > In file included from include/typeinfo:7,
 >                  from ../../egcs/gcc/cp/tinfo.h:4,
 >                  from ../../egcs/gcc/cp/tinfo.cc:29:
 > include/exception:24: Internal compiler error.
 > include/exception:24: Please submit a full bug report to `egcs-bugs@cygnus.com'.
 > 
 > If I change '-fexceptions' to '-fno-exceptions' it doesn't crash.
 > 
 > If I remove -mcoff, (defaulting back to ELF) it doesn't crash.  The core
 > dumps aren't readable by GDB.  Whether this is a GDB problem or an EGCS
 > problem is a different mystery.

With my patch `[980205]: Patch for  wrong mangling (was: Re: [980129]:
bug in COFF's  name mangling?)' submitted   on Mon, 9  Feb 98 12:25:39
+0100 applied I  don't get any  such failures; but  this is  still the
980205 snapshot, I still don't have access to CVS :-(

BTW, did you try the patch?

This is the corresponding part on m88k-motorola-sysv3:

gmake[1]: Entering directory `/home/manfred/work/GNU/egcs-980205-88/gcc'
./xgcc -B./ -O2   -DIN_GCC    -O9 -DUSE_GAS -DUSE_GLD -D__USE_FIXED_PROTOTYPES__ -I./include   -g1  -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -fexceptions   -I. -I../../egcs-980205/gcc -I../../egcs-980205/gcc/config \
  -c ../../egcs-980205/gcc/cp/tinfo.cc
gmake[1]: Leaving directory `/home/manfred/work/GNU/egcs-980205-88/gcc'

And on m68k-motorola-sysv:

gmake[3]: Entering directory `/home/manfred/work/GNU/egcs-980205-68/gcc'
./xgcc -B./ -O2   -DIN_GCC    -O9 -DUSE_GAS -DUSE_GLD -D__USE_FIXED_PROTOTYPES__ -I./include  -DUSE_GAS -g1  -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -fexceptions    -I. -I../../egcs-980205/gcc -I../../egcs-980205/gcc/config \
  -c ../../egcs-980205/gcc/cp/tinfo.cc
gmake[3]: Leaving directory `/home/manfred/work/GNU/egcs-980205-68/gcc'


Manfred

Re: egcs-release branch aborts during bootstrap

[Jeffrey A Law]

  In message <9802130736.AA06633@lts.sel.alcatel.de>you write:
  > On Thu, 12 February 1998, 15:28:22, robertl@dgii.com wrote:
  > 
  >  > [ Manfred, does this fail on your COFF target? ] 
  >  > 
  >  > This command works on the trunk.   It fails on the release tag.
  >  > The sources of the release branch were checked out last night, and
  >  > a 'cvs update -n' in gcc/cp doesn't show any updates pending.  It
  >  > shows up in a make boostrap2 on OpenServer.
  >  > 
  >  > (robertl) rjlhome:/play/egcs-release/native/gcc
  >  > $ cat /tmp/cc
  >  > ./xgcc -B./ -O2   -DIN_GCC    -O2 -g -I./include   -g1 \
  >  >         -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -fexceptions    \
  >  >         -I. -I../../egcs/gcc -I../../egcs/gcc/config -mcoff \
  >  >           -c ../../egcs/gcc/cp/tinfo.cc
  >  > (robertl) rjlhome:/play/egcs-release/native/gcc
  >  > $ sh /tmp/cc
  >  > include/exception: In method `bad_exception::bad_exception()':
  >  > In file included from include/typeinfo:7,
  >  >                  from ../../egcs/gcc/cp/tinfo.h:4,
  >  >                  from ../../egcs/gcc/cp/tinfo.cc:29:
  >  > include/exception:24: Internal compiler error.
  >  > include/exception:24: Please submit a full bug report to `egcs-bugs@cygn
  > us.com'.
  >  > 
  >  > If I change '-fexceptions' to '-fno-exceptions' it doesn't crash.
  >  > 
  >  > If I remove -mcoff, (defaulting back to ELF) it doesn't crash.  The core
  >  > dumps aren't readable by GDB.  Whether this is a GDB problem or an EGCS
  >  > problem is a different mystery.
  > 
  > With my patch `[980205]: Patch for  wrong mangling (was: Re: [980129]:
  > bug in COFF's  name mangling?)' submitted   on Mon, 9  Feb 98 12:25:39
  > +0100 applied I  don't get any  such failures; but  this is  still the
  > 980205 snapshot, I still don't have access to CVS :-(
  > 
  > BTW, did you try the patch?
Where do we stand on this problem?  Is it still aborting in the release
branch?

jeff

Re: egcs-release branch aborts during bootstrap

[Manfred.Hollstein]

On Tue, 3 March 1998, 02:07:14, law@hurl.cygnus.com wrote:

 > 
 >   In message <9802130736.AA06633@lts.sel.alcatel.de>you write:
 >   > On Thu, 12 February 1998, 15:28:22, robertl@dgii.com wrote:
 >   > 
 >   >  > [ Manfred, does this fail on your COFF target? ] 
 >   >  > 
 >   >  > This command works on the trunk.   It fails on the release tag.
 >   >  > The sources of the release branch were checked out last night, and
 >   >  > a 'cvs update -n' in gcc/cp doesn't show any updates pending.  It
 >   >  > shows up in a make boostrap2 on OpenServer.
 >   >  > 
 >   >  > (robertl) rjlhome:/play/egcs-release/native/gcc
 >   >  > $ cat /tmp/cc
 >   >  > ./xgcc -B./ -O2   -DIN_GCC    -O2 -g -I./include   -g1 \
 >   >  >         -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -fexceptions    \
 >   >  >         -I. -I../../egcs/gcc -I../../egcs/gcc/config -mcoff \
 >   >  >           -c ../../egcs/gcc/cp/tinfo.cc
 >   >  > (robertl) rjlhome:/play/egcs-release/native/gcc
 >   >  > $ sh /tmp/cc
 >   >  > include/exception: In method `bad_exception::bad_exception()':
 >   >  > In file included from include/typeinfo:7,
 >   >  >                  from ../../egcs/gcc/cp/tinfo.h:4,
 >   >  >                  from ../../egcs/gcc/cp/tinfo.cc:29:
 >   >  > include/exception:24: Internal compiler error.
 >   >  > include/exception:24: Please submit a full bug report to `egcs-bugs@cygn
 >   > us.com'.
 >   >  > 
 >   >  > If I change '-fexceptions' to '-fno-exceptions' it doesn't crash.
 >   >  > 
 >   >  > If I remove -mcoff, (defaulting back to ELF) it doesn't crash.  The core
 >   >  > dumps aren't readable by GDB.  Whether this is a GDB problem or an EGCS
 >   >  > problem is a different mystery.
 >   > 
 >   > With my patch `[980205]: Patch for  wrong mangling (was: Re: [980129]:
 >   > bug in COFF's  name mangling?)' submitted   on Mon, 9  Feb 98 12:25:39
 >   > +0100 applied I  don't get any  such failures; but  this is  still the
 >   > 980205 snapshot, I still don't have access to CVS :-(
 >   > 
 >   > BTW, did you try the patch?
 > Where do we stand on this problem?  Is it still aborting in the release
 > branch?

Sorry, haven't checked the release branch lately on this machine. I'll start
it soon.

Manfred

Re: egcs-release branch aborts during bootstrap

[Robert Lipe]

>   >  > (robertl) rjlhome:/play/egcs-release/native/gcc
>   >  > $ cat /tmp/cc
>   >  > ./xgcc -B./ -O2   -DIN_GCC    -O2 -g -I./include   -g1 \
>   >  >         -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -fexceptions    \
>   >  >         -I. -I../../egcs/gcc -I../../egcs/gcc/config -mcoff \
>   >  >           -c ../../egcs/gcc/cp/tinfo.cc
>   >  > (robertl) rjlhome:/play/egcs-release/native/gcc
>   >  > $ sh /tmp/cc
>   >  > include/exception: In method `bad_exception::bad_exception()':
>   >  > In file included from include/typeinfo:7,
>   >  >                  from ../../egcs/gcc/cp/tinfo.h:4,
>   >  >                  from ../../egcs/gcc/cp/tinfo.cc:29:
>   >  > include/exception:24: Internal compiler error.
>   >  > include/exception:24: Please submit a full bug report to `egcs-bugs@cygn

> Where do we stand on this problem?  Is it still aborting in the release
> branch?

I've been a slacker and haven't been following the release branch very
closely.

I just ran /tmp/cc and watched it abort.  I did a cvs update followed
by a bootstrap.  I'm still waiting for the f77 stuff to build, but it's
definitely made it past the offending code.  I reran /tmp/cc and it
didn't core.

So I'd say whatever it was went away.

-- 
Robert Lipe       http://www.dgii.com/people/robertl       robertl@dgii.com

Re: egcs-release branch aborts during bootstrap

[Manfred.Hollstein]

On Tue, 3 March 1998, 10:48:14, Manfred.Hollstein@ks.sel.alcatel.de wrote:

 > On Tue, 3 March 1998, 02:07:14, law@hurl.cygnus.com wrote:
 > 
 >  > 
 >  >   In message <9802130736.AA06633@lts.sel.alcatel.de>you write:
 >  >   > On Thu, 12 February 1998, 15:28:22, robertl@dgii.com wrote:
 >  >   > 
 >  >   >  > [ Manfred, does this fail on your COFF target? ] 
 >  >   >  > 
 >  >   >  > This command works on the trunk.   It fails on the release tag.
 >  >   >  > The sources of the release branch were checked out last night, and
 >  >   >  > a 'cvs update -n' in gcc/cp doesn't show any updates pending.  It
 >  >   >  > shows up in a make boostrap2 on OpenServer.
 >  >   >  > 
 >  >   >  > (robertl) rjlhome:/play/egcs-release/native/gcc
 >  >   >  > $ cat /tmp/cc
 >  >   >  > ./xgcc -B./ -O2   -DIN_GCC    -O2 -g -I./include   -g1 \
 >  >   >  >         -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED -fexceptions    \
 >  >   >  >         -I. -I../../egcs/gcc -I../../egcs/gcc/config -mcoff \
 >  >   >  >           -c ../../egcs/gcc/cp/tinfo.cc
 >  >   >  > (robertl) rjlhome:/play/egcs-release/native/gcc
 >  >   >  > $ sh /tmp/cc
 >  >   >  > include/exception: In method `bad_exception::bad_exception()':
 >  >   >  > In file included from include/typeinfo:7,
 >  >   >  >                  from ../../egcs/gcc/cp/tinfo.h:4,
 >  >   >  >                  from ../../egcs/gcc/cp/tinfo.cc:29:
 >  >   >  > include/exception:24: Internal compiler error.
 >  >   >  > include/exception:24: Please submit a full bug report to `egcs-bugs@cygn
 >  >   > us.com'.
 >  >   >  > 
 >  >   >  > If I change '-fexceptions' to '-fno-exceptions' it doesn't crash.
 >  >   >  > 
 >  >   >  > If I remove -mcoff, (defaulting back to ELF) it doesn't crash.  The core
 >  >   >  > dumps aren't readable by GDB.  Whether this is a GDB problem or an EGCS
 >  >   >  > problem is a different mystery.
 >  >   > 
 >  >   > With my patch `[980205]: Patch for  wrong mangling (was: Re: [980129]:
 >  >   > bug in COFF's  name mangling?)' submitted   on Mon, 9  Feb 98 12:25:39
 >  >   > +0100 applied I  don't get any  such failures; but  this is  still the
 >  >   > 980205 snapshot, I still don't have access to CVS :-(
 >  >   > 
 >  >   > BTW, did you try the patch?
 >  > Where do we stand on this problem?  Is it still aborting in the release
 >  > branch?
 > 
 > Sorry, haven't checked the release branch lately on this machine. I'll start
 > it soon.

The bootstrap just finished without any problems (only some NFS server
crashes due to I don't know - they are  at least not related with this
particular configuration!).

Now the runtime  libs have to be built  (around 4 hours) and then make
check will take   another 10 hours.  It's not  this  machine that's so
slooooow, it's the NFS on this machine :-(

I'll keep you informed.

manfred

Re: egcs-release branch aborts during bootstrap

[Manfred Hollstein]

On Thu, 5 March 1998, 16:27:51, Manfred.Hollstein@ks.sel.alcatel.de wrote:

 [snip]
 > The bootstrap just finished without any problems (only some NFS server
 > crashes due to I don't know - they are  at least not related with this
 > particular configuration!).
 > 
 > Now the runtime  libs have to be built  (around 4 hours) and then make
 > check will take   another 10 hours.  It's not  this  machine that's so
 > slooooow, it's the NFS on this machine :-(
 > 
 > I'll keep you informed.

It  took  longer than I'd  expected  :-( But,  results are pretty much
encouraging  (I   just  sent  a   report  to   egcs,  look   for  `CVS
egcs-1.0.2-980303: results on m88k-motorola-sysv3 haifa disabled')!

From my point of view, we can close this problem.

Thanks

manfred