On Thu, Jan 23, 2020 at 12:39:34AM +0100, Jakub Jelinek wrote:
> I don't see how this could have been tested on x86_64-linux, it breaks there
> a lot of tests (ditto i686-linux). My wild guess is that it changes the
> default dg-do action for tests that don't have any, which in tree-ssa.exp
> used to be dg-do compile but now it tries to link them or run them.
> But it seems like it even affects the dg-do default for tests later in other
> *.exp files, e.g. various i386.exp tests, some ubsan.exp etc.
> What tree-ssa tests were the reason for this change?
> I see
> pr88497-1.c:/* { dg-require-effective-target vect_double } */
> pr88497-2.c:/* { dg-require-effective-target vect_float } */
> pr88497-3.c:/* { dg-require-effective-target vect_int } */
> pr88497-4.c:/* { dg-require-effective-target vect_int } */
> pr88497-5.c:/* { dg-require-effective-target vect_int } */
> is that just those? Or the tests that use vect_cmdline_needed effective
> target?
The following patch fixes those, though I'd really like to understand what
was the problem (which tests and why) for you.
2020-01-23 Jakub Jelinek <ja...@redhat.com>
* gcc.dg/tree-ssa/tree-ssa.exp: Save and restore
dg-do-what-default around check_vect_support_and_set_flags call.
--- gcc/testsuite/gcc.dg/tree-ssa/tree-ssa.exp.jj 2020-01-22
22:37:04.523532825 +0100
+++ gcc/testsuite/gcc.dg/tree-ssa/tree-ssa.exp 2020-01-23 00:42:17.530766641
+0100
@@ -28,8 +28,13 @@ set DEFAULT_VECTCFLAGS ""
global EFFECTIVE_TARGETS
set EFFECTIVE_TARGETS ""
+global dg-do-what-default
+set save-dg-do-what-default ${dg-do-what-default}
+
check_vect_support_and_set_flags
+set dg-do-what-default ${save-dg-do-what-default}
+
# If a testcase doesn't have special options, use these.
global DEFAULT_CFLAGS
if ![info exists DEFAULT_CFLAGS] then {
Jakub
