Bug 23676 - [4.1 Regression] ICE: "missing REG_EH_REGION note in the end of bb"
Summary: [4.1 Regression] ICE: "missing REG_EH_REGION note in the end of bb"
Status: RESOLVED FIXED
Alias: None
Product: gcc
Classification: Unclassified
Component: middle-end (show other bugs)
Version: 4.1.0
: P2 normal
Target Milestone: 4.1.0
Assignee: Richard Henderson
URL:
Keywords: build, EH, ice-checking, ice-on-valid-code, wrong-code
: 23696 23712 (view as bug list)
Depends on:
Blocks:
 
Reported: 2005-09-01 11:44 UTC by Andrew Pinski
Modified: 2005-09-03 11:49 UTC (History)
4 users (show)

See Also:
Host:
Target: powerpc-*-*
Build:
Known to work:
Known to fail:
Last reconfirmed: 2005-09-01 16:08:38

Attachments
testcase (22.01 KB, text/plain)
2005-09-02 22:31 UTC, Kazumoto Kojima 		Details
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Andrew Pinski 2005-09-01 11:44:01 UTC 
The following is reduced from the report here:
http://gcc.gnu.org/ml/gcc-regression/2005-09/msg00002.html
cc1plus -fnon-call-exceptions t.cc is all that needed to reproduce the bug.

----
struct JvSynchronize {
  ~JvSynchronize ();
};
struct _Jv_Field {
  char* addr;
};
_Jv_Field * resolve_pool_entry (void);
void * _Jv_ResolvePoolEntry (void)
{
  JvSynchronize sync;
  return (resolve_pool_entry ())->addr;
}
-----
Comment 1 Andrew Pinski 2005-09-01 12:01:25 UTC 
A little more reduced:
struct JvSynchronize {
  ~JvSynchronize ();
};
char* * resolve_pool_entry (void);
void * _Jv_ResolvePoolEntry (void)
{
  JvSynchronize sync;
  return *(resolve_pool_entry ());
}
Comment 2 Richard Henderson 2005-09-01 16:08:37 UTC 
Also reprodicible on ppc-linux.
Comment 3 GCC Commits 2005-09-01 23:35:29 UTC 
Subject: Bug 23676

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	rth@gcc.gnu.org	2005-09-01 23:35:20

Modified files:
	gcc            : ChangeLog reload1.c rtlanal.c 

Log message:
	PR 23676
	* reload1.c (reload_as_needed): Check !CALL_P before calling
	fixup_eh_region_note.
	* rtlanal.c (may_trap_p): SUBREG by itself cannot trap.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.9878&r2=2.9879
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/reload1.c.diff?cvsroot=gcc&r1=1.480&r2=1.481
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/rtlanal.c.diff?cvsroot=gcc&r1=1.218&r2=1.219
Comment 4 Richard Henderson 2005-09-01 23:38:04 UTC 
Fixed.
Comment 5 Andrew Pinski 2005-09-02 15:53:32 UTC 
*** Bug 23696 has been marked as a duplicate of this bug. ***
Comment 6 Kazumoto Kojima 2005-09-02 22:31:19 UTC 
Created attachment 9654 [details]
testcase

There is yet another "missing REG_EH_REGION note in the end of bb"
ICE for sh4-unknown-linux-gnu during libjava build.  I've attached
a reduced preprocessed file, though it could be a new PR.
Comment 7 Andrew Pinski 2005-09-02 22:52:48 UTC 
(In reply to comment #6)
> Created an attachment (id=9654)
> testcase
> 
> There is yet another "missing REG_EH_REGION note in the end of bb"
> ICE for sh4-unknown-linux-gnu during libjava build.  I've attached
> a reduced preprocessed file, though it could be a new PR.

Please file a new bug, as this is more likely a target bug.
Comment 8 Richard Henderson 2005-09-02 22:54:11 UTC 
Subject: Re:  [4.1 Regression] ICE: "missing REG_EH_REGION note in the end of bb"

On Fri, Sep 02, 2005 at 10:31:21PM -0000, kkojima at gcc dot gnu dot org wrote:
> There is yet another "missing REG_EH_REGION note in the end of bb"
> ICE for sh4-unknown-linux-gnu during libjava build.  I've attached
> a reduced preprocessed file, though it could be a new PR.

It's a different problem.  According to the backtrace, this happens
in optimize_mode_switching, and not in reload at all.  I'll look at
it, but please file it as a new bug.


r~
Comment 9 Kazumoto Kojima 2005-09-02 23:35:42 UTC 
I've filed a PR 23706 for it.
Comment 10 Andrew Pinski 2005-09-03 11:49:50 UTC 
*** Bug 23712 has been marked as a duplicate of this bug. ***