[Bug libgomp/106894] [13 regression] multiple libgomp failures after r13-2545-g9f2fca56593a2b

["jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106894

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Ah, sorry about icv-6.c, I did debug just one of the testcases and after fixing
it tested all the mentioned tests and all passed, so thought it is fixed.
I can now reproduce it though.
The thing is,
make check
or
make check RUNTESTFLAGS="c.exp='icv-6.c' c++.exp='icv-6.c'"
in libgomp obj dir work fine, but
make -j32 -k check RUNTESTFLAGS="c.exp='icv-6.c' c++.exp='icv-6.c'"
fails.
The thing is that the testcase as written relies on OMP_NUM_THREADS not being
set in environment (as it takes priority over OMP_NUM_THREADS_ALL for the
host).
So, if either a user has OMP_NUM_THREADS=42 in the environment by himself, or
when doing make check with -jN, we trigger:
          if test $$num_cpus -gt 8 && test -z "$$OMP_NUM_THREADS"; then \
            OMP_NUM_THREADS=8; export OMP_NUM_THREADS; \
            echo @@@ libgomp OMP_NUM_THREADS adjusted to 8 because of parallel
make check and too many CPUs; \
          fi; \
in libgomp/testsuite/Makefile.am and so the test fails.

So, I think we need something like:
--- libgomp/testsuite/libgomp.c-c++-common/icv-6.c.jj   2022-09-08
20:22:07.903182970 +0200
+++ libgomp/testsuite/libgomp.c-c++-common/icv-6.c      2022-09-09
16:49:29.442195701 +0200
@@ -17,6 +17,7 @@

 #include <omp.h>
 #include <stdlib.h>
+#include <string.h>

 int
 main ()
@@ -25,21 +26,28 @@ main ()
   int chunk_size;
   omp_get_schedule(&kind, &chunk_size);

-  if (omp_get_max_teams () != 42
-      || !omp_get_dynamic ()
-      || kind != 3 || chunk_size != 4
-      || omp_get_teams_thread_limit () != 44
-      || omp_get_thread_limit () != 45
-      || omp_get_max_threads () != 46
-      || omp_get_proc_bind () != omp_proc_bind_spread
-      || omp_get_max_active_levels () != 47)
+  if ((!getenv ("OMP_NUM_TEAMS") && omp_get_max_teams () != 42)
+      || (!getenv ("OMP_DYNAMIC") && !omp_get_dynamic ())
+      || (!getenv ("OMP_SCHEDULE") && (kind != 3 || chunk_size != 4))
+      || (!getenv ("OMP_TEAMS_THREAD_LIMIT") && omp_get_teams_thread_limit ()
!= 44)
+      || (!getenv ("OMP_THREAD_LIMIT") && omp_get_thread_limit () != 45)
+      || (!getenv ("OMP_NUM_THREADS") && omp_get_max_threads () != 46)
+      || (!getenv ("OMP_PROC_BIND") && omp_get_proc_bind () !=
omp_proc_bind_spread)
+      || (!getenv ("OMP_MAX_ACTIVE_LEVELS") && omp_get_max_active_levels () !=
47))
     abort ();

   int num_devices = omp_get_num_devices () > 3 ? 3 : omp_get_num_devices ();
-  for (int i=0; i < num_devices; i++)
+  for (int i = 0; i < num_devices; i++)
+    {
+      char name[sizeof ("OMP_NUM_TEAMS_DEV_0")];
+      strcpy (name, "OMP_NUM_TEAMS_DEV_0");
+      name[sizeof ("OMP_NUM_TEAMS_DEV_0") - 2] = '0' + i;
+      if (getenv (name))
+       continue;
     #pragma omp target device (i)
       if (omp_get_max_teams () != 43)
        abort ();
+    }

   return 0;
 }
or if we say users are on their own if they make check with any OMP_*
environment variables,
then at least do that for OMP_NUM_THREADS because we set it ourselves.