]> git.ipfire.org Git - thirdparty/gcc.git/commit
testsuite: Improve unlimited_polymorphic_3.f03 [PR103662]
authorJakub Jelinek <jakub@redhat.com>
Tue, 26 Apr 2022 07:17:32 +0000 (09:17 +0200)
committerJakub Jelinek <jakub@redhat.com>
Tue, 26 Apr 2022 07:17:32 +0000 (09:17 +0200)
commit22c24ba48a20a4944a50cca06449debed7d1b3f4
treec5df1efe401ea2bfeb453f55124203ae29c90325
parent57a957cb71c004de80b0fd30c8db3cc67576e0ce
testsuite: Improve unlimited_polymorphic_3.f03 [PR103662]

On Mon, Apr 25, 2022 at 01:38:25PM +0200, Mikael Morin wrote:
> I have just pushed the attached fix for two UNRESOLVED checks at -O0 that I
> hadn’t seen.

I don't like forcing of DSE in -O0 compilation, wouldn't it be better
to just not check the dse dump at -O0 like in the following patch?

Even better would be to check that the z._data = stores are both present
in *.optimized dump, but that doesn't really work at -O2 or above because
we inline the functions and optimize it completely away (both the stores
and corresponding reads).

The first hunk is needed so that __OPTIMIZE__ effective target works in
Fortran testsuite, otherwise one gets a pedantic error and __OPTIMIZE__
is considered not to match at all.

2022-04-26  Jakub Jelinek  <jakub@redhat.com>

PR fortran/103662
* lib/target-supports.exp (check_effective_target___OPTIMIZE__): Add
a var definition to avoid pedwarn about empty translation unit.
* gfortran.dg/unlimited_polymorphic_3.f03: Remove -ftree-dse from
dg-additional-options, guard scan-tree-dump-not directives on
__OPTIMIZE__ target.
gcc/testsuite/gfortran.dg/unlimited_polymorphic_3.f03
gcc/testsuite/lib/target-supports.exp