Make the internal bot more resilient to timeouts
Change-Id: I0b6d63062b01cb81ee804457045cb3ef2e7019d3
diff --git a/tools/internal_test.py b/tools/internal_test.py
index 182ccf8..81ccd98 100755
--- a/tools/internal_test.py
+++ b/tools/internal_test.py
@@ -334,8 +334,14 @@
log('Running with hash: %s' % git_hash)
exitcode = run_once(archive=True)
log('Running finished with exit code %s' % exitcode)
- put_magic_file(TESTING_COMPLETE, git_hash)
- delete_magic_file(TESTING)
+ # If the bot timed out or something else triggered the bot to fail, don't
+ # put up the result (it will not be displayed anywhere, and we can't
+ # remove the magic file if the bot cleaned up).
+ if get_magic_file_exists(TESTING):
+ put_magic_file(TESTING_COMPLETE, git_hash)
+ # There is still a potential race here (we check, bot deletes, we try to
+ # delete) - this is unlikely and we ignore it (restart if it happens).
+ delete_magic_file(TESTING)
time.sleep(PULL_DELAY)
def handle_output(archive, stderr, stdout, exitcode, timed_out, cmd):
