[Bug tree-optimization/26447] New: verify_flow_info failed

[Bug tree-optimization/26447] New: verify_flow_info failed

[aph at gcc dot gnu dot org]

verify_flow_info tripped an internal error.

OK with -O1 but not with -O2

[aph@zorro tmp]$ /home/aph/install/bin/gcj -c -O1 -g -fPIC -findirect-dispatch
-fjni AbstractCommentParser.class
[aph@zorro tmp]$ /home/aph/install/bin/gcj -c -O2 -g -fPIC -findirect-dispatch
-fjni AbstractCommentParser.class
org/eclipse/jdt/internal/compiler/parser/AbstractCommentParser.java: In class
'org.eclipse.jdt.internal.compiler.parser.AbstractCommentParser':
org/eclipse/jdt/internal/compiler/parser/AbstractCommentParser.java: In method
'org.eclipse.jdt.internal.compiler.parser.AbstractCommentParser.commentParse()':
org/eclipse/jdt/internal/compiler/parser/AbstractCommentParser.java:0: error:
control flow in the middle of basic block 32
org/eclipse/jdt/internal/compiler/parser/AbstractCommentParser.java:0: error:
control flow in the middle of basic block 32
org/eclipse/jdt/internal/compiler/parser/AbstractCommentParser.java:0: internal
compiler error: verify_flow_info failed
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.


-- 
           Summary: verify_flow_info failed
           Product: gcc
           Version: 4.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: aph at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] verify_flow_info failed

[aph at gcc dot gnu dot org]

------- Comment #1 from aph at gcc dot gnu dot org  2006-02-23 16:34 -------
Created an attachment (id=10903)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10903&action=view)
File to compile


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] verify_flow_info failed

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2006-02-24 18:59 -------
Well this one class file is not enough to reproduce this failure as I get
errors:
java/util/Comparator.java: In class
'org.eclipse.jdt.internal.compiler.parser.AbstractCommentParser':
java/util/Comparator.java: In constructor
'(org.eclipse.jdt.internal.compiler.parser.Parser)':
java/util/Comparator.java:114: error: cannot find file for class
org.eclipse.jdt.internal.compiler.parser.Scanner


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] verify_flow_info failed

[aph at gcc dot gnu dot org]

------- Comment #3 from aph at gcc dot gnu dot org  2006-02-27 17:22 -------
I wouldn't expect to see such errors.

Are you sure you used -findirect-dispatch ?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] verify_flow_info failed

[pinskia at gcc dot gnu dot org]

------- Comment #4 from pinskia at gcc dot gnu dot org  2006-02-27 17:23 -------
Oh, I missed that, woops.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE

[pinskia at gcc dot gnu dot org]

------- Comment #5 from pinskia at gcc dot gnu dot org  2006-02-27 17:32 -------
Trying to get a reduced C++ testcase, load PRE is causing the ICE.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org
 GCC target triplet|x86_64-unknown-linux-gnu    |x86_64-linux-gnu
           Keywords|                            |EH, ice-on-valid-code
            Summary|verify_flow_info failed     |[4.2 Regression]
                   |                            |verify_flow_info failed,
                   |                            |load PRE
   Target Milestone|---                         |4.2.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE

[pinskia at gcc dot gnu dot org]

------- Comment #6 from pinskia at gcc dot gnu dot org  2006-02-27 17:41 -------
This the best I can get but I never can get an ICE:
int f(int a, int *b, int *c, int *d)throw()
{
 // try {
  int e = *c;
  if (e!=0)
    *b = 1;
  return *c+*d;
 // } catch(...)
 /* {
    return 0;
  }*/
}
----
Compile at -O2 -fnon-call-exceptions.
We do get:
  #   VUSE <TMT.4_14(ab)>;
  D.2388_8 = *c_1;
  pretmp.21_3 = D.2388_8;

Which looks wrong but I cannot get the ICE.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2006-02-27 17:41:37
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE

[dberlin at dberlin dot org]

------- Comment #7 from dberlin at gcc dot gnu dot org  2006-02-27 19:45 -------
Subject: Bug 26447

Give this patch a try and let me know if it works for you


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE

[dberlin at dberlin dot org]

------- Comment #8 from dberlin at gcc dot gnu dot org  2006-02-27 19:46 -------
Subject: Bug 26447

Whoops, forgot patch


------- Comment #9 from dberlin at gcc dot gnu dot org  2006-02-27 19:46 -------
Created an attachment (id=10923)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10923&action=view)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE

[aph at gcc dot gnu dot org]

------- Comment #10 from aph at gcc dot gnu dot org  2006-03-02 10:31 -------
No, that patch doesn't help.  Still get the same result at -O2:

[aph@zorro eclipse]$ /home/aph/gcc/install/bin/gcj -c -O2 -g -fPIC
-findirect-dispatch -fjni AbstractCommentParser.class
org/eclipse/jdt/internal/compiler/parser/AbstractCommentParser.java: In class
'org.eclipse.jdt.internal.compiler.parser.AbstractCommentParser':
org/eclipse/jdt/internal/compiler/parser/AbstractCommentParser.java: In method
'org.eclipse.jdt.internal.compiler.parser.AbstractCommentParser.commentParse()':
org/eclipse/jdt/internal/compiler/parser/AbstractCommentParser.java:0: error:
control flow in the middle of basic block 32
org/eclipse/jdt/internal/compiler/parser/AbstractCommentParser.java:0: error:
control flow in the middle of basic block 32
org/eclipse/jdt/internal/compiler/parser/AbstractCommentParser.java:0: internal
compiler error: verify_flow_info failed
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE

[pinskia at gcc dot gnu dot org]

------- Comment #11 from pinskia at gcc dot gnu dot org  2006-03-06 15:24 -------
I might look into fix this later this week,  the problem is the creating of
loads which could cause an trap/exception but not putting them into different
BB's.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[pinskia at gcc dot gnu dot org]

------- Comment #12 from pinskia at gcc dot gnu dot org  2006-05-02 15:19 -------
PR 27389 looks like a reduced testcase.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |27389


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[mckinlay at redhat dot com]

------- Comment #13 from mckinlay at redhat dot com  2006-05-02 17:32 -------
*** Bug 27389 has been marked as a duplicate of this bug. ***


-- 
Bug 26447 depends on bug 27389, which changed state.

Bug 27389 Summary: [4.2 Regression] java: verify_flow_info fails
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27389

           What    |Old Value                   |New Value
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |DUPLICATE

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[rguenth at gcc dot gnu dot org]

------- Comment #14 from rguenth at gcc dot gnu dot org  2006-05-03 08:10 -------
This is a blocker for me.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |blocker


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[rguenth at gcc dot gnu dot org]

------- Comment #15 from rguenth at gcc dot gnu dot org  2006-05-03 10:00 -------
Btw, can someone java-capable reduce the testase

import javax.xml.stream.XMLStreamException;
import javax.xml.stream.XMLStreamWriter;
public class StAXWriter
{
  XMLStreamWriter writer;
  int indent = 0;
  public void writeEnd(boolean wasEmpty)
  {
    try
      {
        indent -= 2;
        for (int i = 0; i < indent; i++)
          writer.writeCharacters(" ");
      }
    catch (XMLStreamException xmlse)
      {
      }
  }
};

to something that does not require the imports?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[aph at gcc dot gnu dot org]

------- Comment #16 from aph at gcc dot gnu dot org  2006-05-03 10:11 -------
I can certainly do that, but it doesn't fail on HEAD.  Are you sure you really
want it?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[rguenth at gcc dot gnu dot org]

------- Comment #17 from rguenth at gcc dot gnu dot org  2006-05-03 10:14 -------
It fails for me on head with -O2 -findirect-dispatch:

trunk-g/gcc> ./jc1 -quiet -O2 -findirect-dispatch StAXWriter.java
-fbootclasspath=../i686-pc-linux-gnu/libjava/classpath/lib
StAXWriter.java: In class 'StAXWriter':
StAXWriter.java: In method 'StAXWriter.writeEnd(boolean)':
StAXWriter.java:8: error: control flow in the middle of basic block 3
StAXWriter.java:8: internal compiler error: verify_flow_info failed
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[aph at gcc dot gnu dot org]

------- Comment #18 from aph at gcc dot gnu dot org  2006-05-03 10:22 -------
gcj -findirect-dispatch doesn't work with .java files, only with .class files,
so I didn't try it.

class XMLStreamWriter 
{
  void writeCharacters(String s) {}
}

class XMLStreamException extends Exception {}

public class StAXWriter
{
  XMLStreamWriter writer;
  int indent = 0;
  public void writeEnd(boolean wasEmpty)
  {
    try
      {
        indent -= 2;
        for (int i = 0; i < indent; i++)
          writer.writeCharacters(" ");
      }
    catch (XMLStreamException xmlse)
      {
      }
  }
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[rguenth at gcc dot gnu dot org]

------- Comment #19 from rguenth at gcc dot gnu dot org  2006-05-03 10:36 -------
Thanks!

So, the problem is that PRE inserts (and later realifies) fake stores in basic
blocks with abnormal control flow.  It would need to do the insertion on the
outgoing edges in this case, or deal with splitting the block and adjusting EH
info at realifying point.  But the much easier solution is to just punt on
basic blocks that have abnormal control flow.  Like

Index: tree-ssa-pre.c
===================================================================
*** tree-ssa-pre.c      (revision 113493)
--- tree-ssa-pre.c      (working copy)
*************** insert_fake_stores (void)
*** 3216,3221 ****
--- 3216,3229 ----
    FOR_ALL_BB (block)
      {
        block_stmt_iterator bsi;
+       int i;
+       /* Do not insert fake stores in basic blocks that have
+        abnormal control flow.  */
+       for (i=0; i<EDGE_COUNT (block->succs); ++i)
+       if (EDGE_SUCC (block, i)->flags & EDGE_COMPLEX)
+         break;
+       if (i != EDGE_COUNT (block->succs))
+       continue;
        for (bsi = bsi_start (block); !bsi_end_p (bsi); bsi_next (&bsi))
        {
          tree stmt = bsi_stmt (bsi);

Danny?


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dberlin at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[rguenth at gcc dot gnu dot org]

------- Comment #20 from rguenth at gcc dot gnu dot org  2006-05-03 12:09 -------
Created an attachment (id=11366)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=11366&action=view)
patch

Another patch that implements the suggested basic block splitting by
re-inserting
on the fallthrough edge in realify_fake_stores.

2006-05-03  Richard Guenther  <rguenther@suse.de>

        PR tree-optimization/26447
        * tree-ssa-pre.c (insert_fake_stores): Do not insert fake
        stores in a basic block that contains abnormal control flow
        but more than just a fallthrough edge.
        (realify_fake_stores): For basic blocks with abnormal control
        flow, re-insert the store on the fallthrough edge.

bootstrap/testing in progress.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[rguenth at gcc dot gnu dot org]

------- Comment #22 from rguenth at gcc dot gnu dot org  2006-05-04 07:45 -------
Subject: Bug 26447

Author: rguenth
Date: Thu May  4 07:44:37 2006
New Revision: 113517

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=113517
Log:
2006-05-04  Richard Guenther  <rguenther@suse.de>

        PR tree-optimization/26447
        * tree-ssa-pre.c (realify_fake_stores): For necessary loads
        produce SSA_NAME copies before the store stmt to avoid
        breaking exception handling.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/tree-ssa-pre.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[rguenth at gcc dot gnu dot org]

------- Comment #21 from rguenth at gcc dot gnu dot org  2006-05-04 07:45 -------
Fixed.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447

[Bug tree-optimization/26447] [4.2 Regression] verify_flow_info failed, load PRE with java program

[patchapp at dberlin dot org]

------- Comment #23 from patchapp at dberlin dot org  2006-05-15 19:32 -------
Subject: Bug number PR26447

A patch for this bug has been added to the patch tracker.
The mailing list url for the patch is
http://gcc.gnu.org/ml/gcc-patches/2006-05/msg00138.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26447