This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: tree-ssa internal compiler error
- From: law at redhat dot com
- To: Andrew Pinski <pinskia at physics dot uc dot edu>
- Cc: Diego Novillo <dnovillo at redhat dot com>, Jeffrey dot S dot Whitaker at noaa dot gov, "gcc at gcc dot gnu dot org" <gcc at gcc dot gnu dot org>
- Date: Fri, 01 Aug 2003 12:08:45 -0600
- Subject: Re: tree-ssa internal compiler error
- Reply-to: law at redhat dot com
In message <4C1AFA8E-C360-11D7-AC2E-000393A6D2F2@physics.uc.edu>, Andrew Pinski
writes:
>
>On Thursday, Jul 31, 2003, at 09:58 US/Eastern, Diego Novillo wrote:
>
>> On Tue, 2003-07-29 at 08:10, Jeffrey.S.Whitaker@noaa.gov wrote:
>>> Trying to "make bootstrap" from tree-ssa CVS on 20030728
>>> (powerpc-apple-darwin6.5). This happens with both gcc 3.1 and gcc
>>> 3.3.
>>>
>> I believe that this should be fixed by Jeff Law's patch to bsi_remove
>> (http://gcc.gnu.org/ml/gcc-patches/2003-07/msg02917.html).
>
>Nope, I had just tried right after Jeff Law's patch went in (Thu Jul 31
>05:08:18 UTC 2003)
>and I get the same error as before (and as Jeffrey Whitaker and Daniel
>Berlin get):
>/Volumes/UFS_Partition/pinskia/src/gccssa/src/libbanshee/engine/hash.c:
>In function `hash_table_remove':
>/Volumes/UFS_Partition/pinskia/src/gccssa/src/libbanshee/engine/
>hash.c:219: internal compiler error: tree check: expected ssa_name,
>have var_decl in process_worklist, at tree-ssa-dce.c:417
>Please submit a full bug report,
>with preprocessed source if appropriate.
>See <URL:http://gcc.gnu.org/bugs.html> for instructions.
I doubt my change would have fixed this kind of bug. This sounds
more like something is mixing virtual and real operands.
It's _possible_ that propagation into virtual operands could trigger
this kind of failure -- though that would be an indication that
something else is buggered up before we copy propagate into virtual
operands.
I'd be very interested in seeing what the statement in question looks like,
along with its virtual operands. I believe you can get that via
dump_generic_stmt (i) "i" is a statement in the context where we're
aborting.
jeff