[Bug lto/65302] [5 Regression] LTO: ICE internal compiler error: verify_flow_info failed

[Bug lto/65302] [5 Regression] LTO: ICE internal compiler error: verify_flow_info failed

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65302

--- Comment #1 from Tobias Burnus <burnus at gcc dot gnu.org> ---
Created attachment 34939
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=34939&action=edit
two.ii

[Bug lto/65302] New: [5 Regression] LTO: ICE internal compiler error: verify_flow_info failed

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65302

            Bug ID: 65302
           Summary: [5 Regression] LTO: ICE internal compiler error:
                    verify_flow_info failed
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: lto
          Assignee: unassigned at gcc dot gnu.org
          Reporter: burnus at gcc dot gnu.org
                CC: hubicka at gcc dot gnu.org

Created attachment 34938
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=34938&action=edit
one.ii

Follow up to PR65276. The following code fails with an ICE. The ICE is gone if
one removes the "-O0" from the LTO stage:

  g++ -c -flto -O2 one.ii two.ii
  gcc -r -nostdlib -O0 one.o two.o

fails with

one.ii: In member function ‘Init’:
one.ii:93:6: error: BB 5 can not throw but has an EH edge
 void ZRubberBand2P::Init () {
      ^
one.ii:93:6: internal compiler error: verify_flow_info failed
0x690a8b verify_flow_info()
        ../../gcc/cfghooks.c:280


Both PR65276 and this PR are for the same big code; I fear that neither the
fixed PR65276 nor this PR are fully representative for the big code as the
latter should be compiled with some "-O" while the two PRs involve -O0. Thus,
expect another follow-up PR once this PR is fixed. The compilation of the big
program fails with an ICE in add_symbol_to_partition_1, at
lto/lto-partition.c:158.
>From gcc-bugs-return-479180-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Tue Mar 03 16:02:08 2015
Return-Path: <gcc-bugs-return-479180-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 27636 invoked by alias); 3 Mar 2015 16:02:08 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 27366 invoked by uid 48); 3 Mar 2015 16:02:05 -0000
From: "law at redhat dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug rtl-optimization/64081] [5 Regression] r217827 prevents RTL loop unroll
Date: Tue, 03 Mar 2015 16:02:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: rtl-optimization
X-Bugzilla-Version: 5.0
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: normal
X-Bugzilla-Who: law at redhat dot com
X-Bugzilla-Status: UNCONFIRMED
X-Bugzilla-Priority: P2
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 5.0
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields: priority
Message-ID: <bug-64081-4-ZM48TYt1OQ@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-64081-4@http.gcc.gnu.org/bugzilla/>
References: <bug-64081-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2015-03/txt/msg00324.txt.bz2
Content-length: 286

https://gcc.gnu.org/bugzilla/show_bug.cgi?idd081

Jeffrey A. Law <law at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2

[Bug lto/65302] [5 Regression] LTO: ICE internal compiler error: verify_flow_info failed

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65302

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |5.0

[Bug lto/65302] [5 Regression] LTO: ICE internal compiler error: verify_flow_info failed

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65302

--- Comment #3 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
Started with r218767:

commit 9e7bd3ae5774b5b9d383588f42d8edab0003a9bf
Author: hubicka <hubicka@138bc75d-0d04-0410-961f-82ee72b054a4>
Date:   Mon Dec 15 22:35:20 2014 +0000

        PR lto/64043
        * gcc.dg/lto/20110201-1_0.c: New testcase.

        * tree-streamer.c (preload_common_nodes): Skip preloading
        of main_identifier_node, pid_type and optimization/option nodes.


    git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@218767
138bc75d-0d04-0410-96
1f-82ee72b054a4

[Bug lto/65302] [5 Regression] LTO: ICE internal compiler error: verify_flow_info failed

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65302

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |hubicka at gcc dot gnu.org

--- Comment #4 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
*free_cfg_annotations is bogus name for pass_fixup_cfg, but it is precisely
what the pass should do - clear EH edges if they are no longer needed.

I will take a look.

[Bug lto/65302] [5 Regression] LTO: ICE internal compiler error: verify_flow_info failed

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65302

--- Comment #5 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
BB 5 is:
<bb 5>:
getStringBounds (_4, _11, &D.4106, _9, _8, _7);
goto <bb 7>;

and fixup cfg calls:

8702              if (maybe_clean_eh_stmt (stmt)
8703                  && gimple_purge_dead_eh_edges (bb))

the statement is correctly identified as nonthrowing but edges are removed,
because it is not in EH region.

It is created by:
#0  0x0000000000921eb5 in gimple_copy (stmt=0x7ffff68e0e58) at
../../gcc/gimple.c:1687
#1  0x0000000000f96735 in gimple_ic (icall_stmt=0x7ffff68e0e58,
direct_call=0x7ffff68e6498, prob=8000, count=0, all=0) at
../../gcc/value-prof.c:1439
#2  0x00000000007976f2 in cgraph_edge::redirect_call_stmt_to_callee
(this=0x7ffff6adf7b8) at ../../gcc/cgraph.c:1336
#3  0x000000000155a53a in inline_transform (node=0x7ffff6ae57a8) at
../../gcc/ipa-inline-transform.c:529

later the EH is copied but only if stmt can throw:

  /* Build an EH edge for the direct call if necessary.  */                     
  lp_nr = lookup_stmt_eh_lp (icall_stmt);                                       
  if (lp_nr > 0 && stmt_could_throw_p (dcall_stmt))                             
    {                                                                           
      add_stmt_to_eh_lp (dcall_stmt, lp_nr);                                    
    }                                                                           

I suppose we want to purge EH edges in case we did not copy the EH region.

Index: value-prof.c
===================================================================
--- value-prof.c        (revision 221223)
+++ value-prof.c        (working copy)
@@ -1576,6 +1576,8 @@ gimple_ic (gcall *icall_stmt, struct cgr
                     PHI_ARG_DEF_FROM_EDGE (phi, e_eh));
          }
        }
+  if (!stmt_could_throw_p (dcall_stmt))
+    gimple_purge_dead_eh_edges (dcall_bb);
   return dcall_stmt;
 }

[Bug lto/65302] [5 Regression] LTO: ICE internal compiler error: verify_flow_info failed

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65302

--- Comment #6 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Author: hubicka
Date: Fri Mar  6 16:12:43 2015
New Revision: 221244

URL: https://gcc.gnu.org/viewcvs?rev=221244&root=gcc&view=rev
Log:

    PR ipa/65302
    * value-prof.c (gimple_ic): Pure dead eh edges when needed.
    * g++.dg/lto/pr65302_1.C: New testcase.
    * g++.dg/lto/pr65302_0.C: New testcase.

Added:
    trunk/gcc/testsuite/g++.dg/lto/pr65302_0.C
    trunk/gcc/testsuite/g++.dg/lto/pr65302_1.C
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/value-prof.c

[Bug lto/65302] [5 Regression] LTO: ICE internal compiler error: verify_flow_info failed

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65302

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

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

--- Comment #7 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Fixed.