gcc-4.2-hppa64 bug

gcc-4.2-hppa64 bug

[Kyle McMartin]

John et al,

We're currently hitting a gcc bug on hppa64 with both gcc-4.1 and
gcc-4.2. Symptoms are that a build of the ocfs2 module spins
(apparently) forever, I'm leaving a copy building to see if it
eventually makes forward progress.

It's pretty easily reproduceable by building a modular kernel with the
ocfs2 module enabled. Is there anything I can do to help narrow this
down?

make fs/ocfs2/ocfs2.ko:
	!SMP	SMP
32-bit	 y	 y
64-bit	 y	 fails

The troublesome source file appears to be fs/ocfs2/alloc.o, the rest
seem to build fine.

cheers,
	Kyle

Re: gcc-4.2-hppa64 bug

[Kyle McMartin]

On Wed, Dec 05, 2007 at 12:02:19PM -0500, Kyle McMartin wrote:
> The troublesome source file appears to be fs/ocfs2/alloc.o, the rest
> seem to build fine.
> 

Randolph pointed out the "-da" argument to me, the last pass it makes it
to is alloc.c.160r.shorten.

I'll leave this one going to check for forward progress as well.

cheers,
	Kyle

Re: gcc-4.2-hppa64 bug

[John David Anglin]

> It's pretty easily reproduceable by building a modular kernel with the
> ocfs2 module enabled. Is there anything I can do to help narrow this
> down?

The general process is to determine the compilation command that
fails.  "-da" can be used to determine the last function that's
being compiled.  Then, try to whittle down the testcase to something
that's easier to debug.  That's a good point to submit a GCC PR.

To actually debug the compilation, use "-v" to see the command gcc
uses to run cc1.  Run cc1 under gdb with the arguments shown with
"-v".  However, you probably won't get very far without a version
of cc1 with debug symbols.  It's also easiest to debug cc1 from
stage1 of the GCC bootstrap.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)

Re: gcc-4.2-hppa64 bug

[Aurelien Jarno]

Kyle McMartin a =E9crit :
> John et al,
>=20
> We're currently hitting a gcc bug on hppa64 with both gcc-4.1 and
> gcc-4.2. Symptoms are that a build of the ocfs2 module spins
> (apparently) forever, I'm leaving a copy building to see if it
> eventually makes forward progress.
>=20
> It's pretty easily reproduceable by building a modular kernel with the
> ocfs2 module enabled. Is there anything I can do to help narrow this
> down?
>=20
> make fs/ocfs2/ocfs2.ko:
> 	!SMP	SMP
> 32-bit	 y	 y
> 64-bit	 y	 fails
>=20
> The troublesome source file appears to be fs/ocfs2/alloc.o, the rest
> seem to build fine.

Please note that this bug has been reported sometimes ago in the
upstream bug tracker. See: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D=
31944

Cheers,
Aurelien

--=20
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net


--=20
To UNSUBSCRIBE, email to debian-hppa-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian=
.org

Re: gcc-4.2-hppa64 bug

[John David Anglin]

> Please note that this bug has been reported sometimes ago in the
> upstream bug tracker. See: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31944

I marked this bug as confirmed and a regression.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)