On 2019/5/6 7:35 PM, Jakub Jelinek wrote:
On Mon, May 06, 2019 at 01:19:15PM +0200, Christophe Lyon wrote:
The regression testing for the patch was done on GCC mainline on
powerpc64le-unknown-linux-gnu (Power 9 LE)
with no regressions. Is it OK for trunk and backport to gcc 9 ?
While the posted patch had:
+/* PR tree-optimization/88676 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+/* { dg-final { scan-tree-dump-not " = PHI <" "optimized" } } */
the actually committed patch has:
/* PR tree-optimization/88676 */
/* { dg-do compile } */
/* { dg-options "-O2 -fdump-tree-phiopt1" } */
/* { dg-final { scan-tree-dump-not " = PHI <" "optimized" } } */
Dunno why this changed, if you want it in phiopt1, you need "phiopt1"
in scan-tree-dump-not as well, if you want optimized dump, you need
-fdump-tree-optimized instead.
When I test the code again for submiting the code, I found that this
change only affects the dump file generated by phiopt1,
so I rashly decided to change the test option to dump-tree-phiopt1. If
there is any code change in the future, I will send another patch.
Sorry for the error caused by this.
This new testcase fails on arm and aarch64:
PASS: gcc.dg/tree-ssa/pr88676-2.c (test for excess errors)
UNRESOLVED: gcc.dg/tree-ssa/pr88676-2.c scan-tree-dump-not optimized " = PHI <"
because:
gcc.dg/tree-ssa/pr88676-2.c: dump file does not exist
Can you fix this?
Jakub
