[glib: 10/11] tests: Continue closure-refcount test until all three threads are seen

[Philip Withnall <pwithnall src gnome org>]

commit 1ae54da69b7209d48ef1360ec1ad79bf53f2b99c
Author: Philip Withnall <withnall endlessm com>
Date:   Wed Feb 27 12:26:59 2019 +0000

    tests: Continue closure-refcount test until all three threads are seen
    
    Previously, the test assumed that thread1 and thread2 would be scheduled
    enough to set seen_thread{1,2} by the fact that the test runs for a high
    number of iterations. On some platforms/schedulers, that’s not true,
    which causes the test to spuriously fail.
    
    Fix that by forcing the test to continue iterating until both threads
    are seen. If this takes too long, the Meson test runner timeout will be
    hit and the test will be terminated.
    
    Signed-off-by: Philip Withnall <withnall endlessm com>

 gobject/tests/closure-refcount.c | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)
---
diff --git a/gobject/tests/closure-refcount.c b/gobject/tests/closure-refcount.c
index 37c9637d8..c64c366e1 100644
--- a/gobject/tests/closure-refcount.c
+++ b/gobject/tests/closure-refcount.c
@@ -258,7 +258,7 @@ test_closure_refcount (void)
   TestClosureRefcountData test_data = { 0, };
   GClosure *closure;
   GTest *object;
-  guint i;
+  guint i, n_iterations;
 
   object = g_object_new (G_TYPE_TEST, NULL);
   closure = g_cclosure_new (G_CALLBACK (test_signal_handler), &test_data, destroy_data);
@@ -281,10 +281,21 @@ test_closure_refcount (void)
    * https://gitlab.gnome.org/GNOME/glib/issues/1316
    * aka https://bugs.debian.org/880883 */
 #if defined(__aarch64__) || defined(__arm__)
-  for (i = 0; i < 100000; i++)
+  n_iterations = 100000;
 #else
-  for (i = 0; i < 1000000; i++)
+  n_iterations = 1000000;
 #endif
+
+  /* Run the test for a reasonably high number of iterations, and ensure we
+   * don’t terminate until at least 10000 iterations have completed in both
+   * thread1 and thread2. Even though @n_iterations is high, we can’t guarantee
+   * that the scheduler allocates time fairly (or at all!) to thread1 or
+   * thread2. */
+  for (i = 0;
+       i < n_iterations ||
+       !g_atomic_int_get (&test_data.seen_thread1) ||
+       !g_atomic_int_get (&test_data.seen_thread2);
+       i++)
     {
       test_emissions (object);
       if (i % 10000 == 0)