4 p.m.
Anyone have any good ideas how to work around this?
andy
event-stream.c: In function `menu_move_left':
event-stream.c:3189: internal error--unrecognizable insn:
(insn 256 14 15 (set (reg/v:SI 24)
(minus:SI (reg/v:SI 25)
(const_int 2))) -1 (insn_list 11 (nil))
(nil))
0 0 [main]
C:\Cygwin\usr\lib\gcc-lib\i686-cygwin\2.9-cygwin-990830\cc1
.exe 11344 handle_exceptions: Exception: STATUS_ACCESS_VIOLATION
0 11409 [main] cc1 11344 stackdump: Dumping stack trace to
cc1.exe.stackdu
mp
make: *** [event-stream.o] Error 1
--------------------------------------------------------------
Dr Andy Piper
Senior Consultant Architect, BEA Systems Ltd
5:07 p.m.
Is the code current cvs? May need to rewrite it a bit.
-eric
event-stream.c: In function `menu_move_left':
event-stream.c:3189: internal error--unrecognizable insn:
(insn 256 14 15 (set (reg/v:SI 24)
(minus:SI (reg/v:SI 25)
(const_int 2))) -1 (insn_list 11 (nil))
(nil))
0 0 [main]
C:\Cygwin\usr\lib\gcc-lib\i686-cygwin\2.9-cygwin-990830\cc1
.exe 11344 handle_exceptions: Exception: STATUS_ACCESS_VIOLATION
0 11409 [main] cc1 11344 stackdump: Dumping stack trace to
cc1.exe.stackdu
mp
11:30 p.m.
>>>> "Eric" == Eric Christopher
<echristo(a)cygnus.com> writes:
Eric> Is the code current cvs? May need to rewrite it a bit.
This happened on Linux as well somewhere. I looked at this, and it
seemed that A called B called C, B and C are static, B gets inlined
into A, C gets inlined into B. Any change to gcc's inlining algorithm
(like changing the size of the function) caused the crash to go away.
I couldn't come up with a small test case.
IIRC you can get around this by compiling without optimization, or
perhaps even -fno-inline
Martin
Eric> -eric
>
> event-stream.c: In function `menu_move_left':
> event-stream.c:3189: internal error--unrecognizable insn:
> (insn 256 14 15 (set (reg/v:SI 24)
> (minus:SI (reg/v:SI 25)
> (const_int 2))) -1 (insn_list 11 (nil))
> (nil))
> 0 0 [main]
> C:\Cygwin\usr\lib\gcc-lib\i686-cygwin\2.9-cygwin-990830\cc1
> .exe 11344 handle_exceptions: Exception: STATUS_ACCESS_VIOLATION
> 0 11409 [main] cc1 11344 stackdump: Dumping stack trace to
> cc1.exe.stackdu
> mp
1:52 a.m.
Are you using the Cygwin V1.0 compiler?
If so, I've found that building with just -O (instead of -O2)
fixes it as does installing Mumit's 2.95.2 for v1.0.
Todd
-----Original Message-----
From: owner-xemacs-beta(a)xemacs.org
[mailto:owner-xemacs-beta＠xemacs.org]On Behalf Of Andy Piper
Sent: Thursday, December 02, 1999 11:00 AM
To: xemacs-beta(a)xemacs.org
Subject: gcc brokeness
Anyone have any good ideas how to work around this?
andy
event-stream.c: In function `menu_move_left':
event-stream.c:3189: internal error--unrecognizable insn:
(insn 256 14 15 (set (reg/v:SI 24)
(minus:SI (reg/v:SI 25)
(const_int 2))) -1 (insn_list 11 (nil))
(nil))
0 0 [main]
C:\Cygwin\usr\lib\gcc-lib\i686-cygwin\2.9-cygwin-990830\cc1
.exe 11344 handle_exceptions: Exception: STATUS_ACCESS_VIOLATION
0 11409 [main] cc1 11344 stackdump: Dumping stack trace to
cc1.exe.stackdu
mp
make: *** [event-stream.o] Error 1
--------------------------------------------------------------
Dr Andy Piper
Senior Consultant Architect, BEA Systems Ltd
12:47 p.m.
At 08:52 PM 12/2/99 -0500, Todd Goodman wrote:
Are you using the Cygwin V1.0 compiler?
Yes.
If so, I've found that building with just -O (instead of -O2)
fixes it as does installing Mumit's 2.95.2 for v1.0.
Ok. We need to come up with something that means people can build OOTB with
cygwin 1.0.
andy
--------------------------------------------------------------
Dr Andy Piper
Senior Consultant Architect, BEA Systems Ltd
4:49 a.m.
>>>> "Andy" == Andy Piper
<andy(a)xemacs.org> writes:
Andy> At 08:52 PM 12/2/99 -0500, Todd Goodman wrote:
> Are you using the Cygwin V1.0 compiler?
Andy> Yes.
> If so, I've found that building with just -O (instead of
-O2)
> fixes it as does installing Mumit's 2.95.2 for v1.0.
Andy> Ok. We need to come up with something that means people can build OOTB with
Andy> cygwin 1.0.
Here's the fix that lets me compile with Linux CodeFusion. It is
likely that the same fix will work for Cygwin. Someone (Andy?) please
test this on Cygwin 1.0.
Cygnus and Cygwin guys, you're not off the hook yet. You still need
to fix this serious bug!
Clues:
-O3 gives ICE
-O2 gives ICE
-O gives clean compile
-O2 -fno-strength-reduce gives clean compile
So a reasonable guess is that strength reduction optimizations are
broken.
--- event-stream.c 1999/11/28 18:15:46 1.45.2.14
+++ event-stream.c 1999/12/04 04:44:42
＠＠ -3177,11 +3169,9 ＠＠
int l = level;
widget_value *current;
- while (level >= 3)
- {
- --level;
- lw_pop_menu ();
- }
+ while (level-- >= 3)
+ lw_pop_menu ();
+
menu_move_up ();
current = lw_get_entries (False);
if (l > 2 && current->contents)
＠＠ -3195,11 +3185,9 ＠＠
int l = level;
widget_value *current;
- while (level >= 3)
- {
- --level;
- lw_pop_menu ();
- }
+ while (level-- >= 3)
+ lw_pop_menu ();
+
menu_move_down ();
current = lw_get_entries (False);
if (l > 2 && current->contents)
10:06 p.m.
At 08:49 PM 12/3/99 -0800, Martin Buchholz wrote:
Cygnus and Cygwin guys, you're not off the hook yet. You still
need
to fix this serious bug!
-O2 -fno-strength-reduce gives clean compile
So a reasonable guess is that strength reduction optimizations are
broken.
Well, it looks like these optimizations are also repsonsible for all my X
lossage. Xpm compiled with these crashes XEmacs, Xt compiled with these
just doesn't work at all. So I would say the problem is pretty comprehensive.
I would recommend anybody using cygwin 1.0 compiles with
-fno-strength-reduce for everything.
andy
--------------------------------------------------------------
Dr Andy Piper
Senior Consultant Architect, BEA Systems Ltd
9466
days inactive
9469
days old
6 comments
4 participants
participants (4)
-
Andy Piper -
Eric Christopher -
Martin Buchholz -
Todd Goodman