--- Comment #8 from matz at gcc dot gnu dot org 2007-10-22 15:32 ---
Fixed.
--
matz at gcc dot gnu dot org changed:
What|Removed |Added
Status|NEW
--- Comment #7 from matz at gcc dot gnu dot org 2007-10-22 15:29 ---
Subject: Bug 33855
Author: matz
Date: Mon Oct 22 15:29:17 2007
New Revision: 129551
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=129551
Log:
PR tree-optimization/33855
* tree-ssa-phiopt.c (co
--- Comment #6 from pinskia at gcc dot gnu dot org 2007-10-22 08:56 ---
# cstore.12_30 = VDEF
cstore.12_30 = COMPLEX_EXPR ;
That does not make sense either. How can there be a VDEF on a gimple register
variable?
This is caused by the cselim pass which was added by:
2007-09-10 M
--- Comment #5 from tbm at cyrius dot com 2007-10-22 07:10 ---
Trunk from 20070902 works, 20070916 shows the ICE.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33855
--- Comment #4 from tbm at cyrius dot com 2007-10-22 07:08 ---
When I change the testcase slightly, namely
- vsorc.vqd[indx] = t*0.0fj;
+ vsorc.vqd[indx] = t;
then I get:
(sid)30051:[EMAIL PROTECTED]: ~] /usr/lib/gcc-snapshot/bin/gcc -c -O1
nec2c-main.c
nec2c-main.c: In funct
--- Comment #3 from tbm at cyrius dot com 2007-10-22 07:04 ---
Program received signal SIGSEGV, Segmentation fault.
verify_ssa (check_modified_stmt=1 '\001') at gcc/tree-ssa.c:627
627 if (!IS_EMPTY_STMT (stmt))
(gdb) where
#0 verify_ssa (check_modified_stmt=1 '\001') at gc
--- Comment #2 from tbm at cyrius dot com 2007-10-22 07:04 ---
/* Testcase by Martin Michlmayr <[EMAIL PROTECTED]> */
typedef struct {
int nsant, nvqd;
_Complex long double *vqd;
} vsorc_t;
vsorc_t vsorc;
void foo(int next_job, int ain_num, int iped, long t) {
long double zpnorm;
--- Comment #1 from tbm at cyrius dot com 2007-10-22 07:03 ---
Created an attachment (id=14386)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14386&action=view)
preprocessed source
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33855
