We want to keep the first timeline in an incomplete state as we replace the syncobj's fence; so we need to defer the close() until the end of the test. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103188 Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> Cc: Jason Ekstrand <jason@xxxxxxxxxxxxxx> Cc: Dave Airlie <airlied@xxxxxxxxxx> --- tests/syncobj_wait.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/tests/syncobj_wait.c b/tests/syncobj_wait.c index 5fd37808..2d5328cc 100644 --- a/tests/syncobj_wait.c +++ b/tests/syncobj_wait.c @@ -525,7 +525,7 @@ test_wait_snapshot(int fd, uint32_t test_flags) { struct wait_thread_data wait = { 0 }; uint32_t syncobjs[2]; - int timelines[2] = { -1, -1 }; + int timelines[3] = { -1, -1, -1 }; pthread_t thread; syncobjs[0] = syncobj_create(fd, 0); @@ -570,7 +570,7 @@ test_wait_snapshot(int fd, uint32_t test_flags) * the kernel picks up on the new fence (it shouldn't), we'll get a * timeout. */ - close(timelines[0]); + timelines[2] = timelines[0]; timelines[0] = syncobj_attach_sw_sync(fd, syncobjs[0]); sleep_nsec(SHORT_TIME_NSEC / 5); @@ -589,6 +589,7 @@ test_wait_snapshot(int fd, uint32_t test_flags) close(timelines[0]); close(timelines[1]); + close(timelines[2]); syncobj_destroy(fd, syncobjs[0]); syncobj_destroy(fd, syncobjs[1]); } -- 2.15.0.rc0 _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
