Subject: Re:  [4.4 Regression] natGC.cc:229: internal compiler error: Segmentation fault

need .ii file at least.


On Tue, Jul 8, 2008 at 9:15 PM, pinskia at gcc dot gnu dot org
<gcc-bugzilla@gcc.gnu.org> wrote:
>
>
> --
>
> pinskia at gcc dot gnu dot org changed:
>
>           What    |Removed                     |Added
> ----------------------------------------------------------------------------
>                 CC|                            |dberlin at gcc dot gnu dot
>                   |                            |org
>           Keywords|                            |build, ice-on-valid-code
>            Summary|natGC.cc:229: internal      |[4.4 Regression]
>                   |compiler error: Segmentation|natGC.cc:229: internal
>                   |fault                       |compiler error: Segmentation
>                   |                            |fault
>   Target Milestone|---                         |4.4.0
>
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36766
>
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.
>

Also occurs on hppa-unknown-linux-gnu.

Created attachment 15900 [details]
natGC.ii

The previously attached natGC.ii is from powerpc-apple-darwin9.3.0.

Is this fixed now?

Subject: Re:  [4.4 Regression] natGC.cc:229:
	internal compiler error: Segmentation fault

> Is this fixed now?

No.  Still broken with revision 137996.

Dave

x86_64-linux-gnu (obviously) works for me.

Subject: Re:  [4.4 Regression] natGC.cc:229: internal compiler error: Segmentation fault

I tried i686-darwin and it work for me too.
I'll try cross to ppc-darwin.


On Sun, Jul 20, 2008 at 7:16 AM, rguenth at gcc dot gnu dot org
<gcc-bugzilla@gcc.gnu.org> wrote:
>
>
> ------- Comment #7 from rguenth at gcc dot gnu dot org  2008-07-20 11:16 -------
> x86_64-linux-gnu (obviously) works for me.
>
>
> --
>
> rguenth at gcc dot gnu dot org changed:
>
>           What    |Removed                     |Added
> ----------------------------------------------------------------------------
>  GCC build triplet|x86_64-suse-linux           |
>   GCC host triplet|x86_64-suse-linux           |
>  GCC target triplet|x86_64-suse-linux           |ppc-apple-darwin, hppa-
>                   |                            |linux-gnu
>
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36766
>
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.
>

aehm, triggers on ppc-linux as well. Important key: --enable-java-awt=xlib

Otherwise it does not trigger on a default config for libjava.

Subject: Re:  [4.4 Regression] natGC.cc:229: internal compiler error: Segmentation fault

> ------- Comment #7 from rguenth at gcc dot gnu dot org  2008-07-20 11:16 -------
> x86_64-linux-gnu (obviously) works for me.

Have you tried configuring with --enable-java-awt="xlib"?  Otherwise,
it works for me with suse 11.0.

Dave

--enable-java-awt=xlib makes it break under i686-darwin and i686-linux as well. 

Can you then provide preprocessed source that breaks? ;)

Subject: Re:  [4.4 Regression] natGC.cc:229:
	internal compiler error: Segmentation fault

Attached x86_64 version.  This is the third time of tried to send this.

Dave

Created attachment 15932 [details]
natGC.ii.gz

Created attachment 15933 [details]
preprocessed source from i686-linux

The preprocessed source from i686-linux....

With the first preprocessed source I can reproduce this with -O2 -fnon-call-exceptions on powerpc-linux-gnu, reducing.

Reduced testcase compile with -O1 -fnon-call-exceptions:
struct _Vector_base     {
  ~_Vector_base()       {
   int* _M_start1 = this->_M_start;
  }
  int* _M_start;
};
struct vector  : _Vector_base
{
  vector(){
    int* _M_start1 = this->_M_start;
  }
  ~vector(){
    int* _M_start1 = this->_M_start;
  }
};
void  updateClip() {
  vector* xrectvector = new vector;
  delete xrectvector;
}
--- CUT ---

Subject: Re:  [4.4 Regression] natGC.cc:229: internal compiler error: Segmentation fault

All the blocks we have marked in the bitmap to eh cleanup exist when
we pass them to tree_purge_dead_eh_edges.
tree_purge_dead_eh_edges seems to decide to delete some of them
earlier than it expects, and then crashes when they don't later exist.

On Sun, Jul 20, 2008 at 4:39 PM, rguenth at gcc dot gnu dot org
<gcc-bugzilla@gcc.gnu.org> wrote:
>
>
> --
>
> rguenth at gcc dot gnu dot org changed:
>
>           What    |Removed                     |Added
> ----------------------------------------------------------------------------
>                 CC|                            |rguenth at gcc dot gnu dot
>                   |                            |org
>
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36766
>
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.
>

Fyi, this seems resloved on the tuples branch. Tested on various darwin combinations and on i686-linux.

I can reproduce this on the trunk after the tuples merge (i686).

Ok, to clarify, the bootstrap issue 'has gone', but the reduced test case from #17 still fails. Failure confirmed on i686-apple-darwin r138861.

Created attachment 16191 [details]
gcc44-pr36766.patch

I don't see tree-ssa-dom.c or tree-ssa-pre.c are at fault here (not guessing that some basic block will be purged as unneeded), so it should be gimple_purge_all_eh_edges that needs to be changed to handle this.

Subject: Bug 36766

Author: jakub
Date: Tue Sep  2 19:13:47 2008
New Revision: 139908

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=139908
Log:
	PR tree-optimization/36766
	* tree-cfg.c (gimple_purge_all_dead_eh_edges): Do nothing
	for already removed basic blocks.

	* g++.dg/tree-ssa/pr36766.C: New test.

Added:
    trunk/gcc/testsuite/g++.dg/tree-ssa/pr36766.C
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-cfg.c

Fixed.

Subject: Re:  [4.4 Regression] natGC.cc:229: internal compiler error: Segmentation fault

> Fixed.

Thanks,
Dave