This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: ivtmp problem in ACATS c85006e
- From: Diego Novillo <dnovillo at redhat dot com>
- To: Richard Kenner <kenner at vlsi1 dot ultra dot nyu dot edu>
- Cc: gcc at gcc dot gnu dot org
- Date: Tue, 18 Jan 2005 18:04:20 -0500
- Subject: Re: ivtmp problem in ACATS c85006e
- Organization: Red Hat Canada
- References: <10501182300.AA09690@vlsi1.ultra.nyu.edu>
Richard Kenner wrote:
I keep hoping this will go away, but it doesn't.
Can somebody give me some suggestions as to how to debug it?
c85006e.adb: In function 'C85006E':
c85006e.adb:38: error: Definition in block 206 does not dominate use in block 1108
for SSA_NAME: D.3021_6076in statement:
ivtmp.1749_80845 = &(*D.3022_6077)[5 ...]{lb: D.3021_6076 sz: 1}[5]{lb: 5 sz: 1\};
This is the last remaining ICE for ACATS.
Sometimes happens when some newly exposed symbol does not get fully
renamed. I would start by checking where D.3021_6076 is generated.
To test a hunch, see if this makes the problem go away (I'm hoping it
won't). If it does, would you file a PR and assign it to me?
Thanks. Diego.
--- tree-into-ssa.c 29 Dec 2004 21:45:01 -0000 2.21.2.8
+++ tree-into-ssa.c 18 Jan 2005 23:03:09 -0000
@@ -385,7 +385,7 @@ mark_def_sites (struct dom_walk_data *wa
stmt = bsi_stmt (bsi);
get_stmt_operands (stmt);
- REWRITE_THIS_STMT (stmt) = 0;
+ REWRITE_THIS_STMT (stmt) = 1;
/* If a variable is used before being set, then the variable is live
across a block boundary, so mark it live-on-entry to BB. */