> I run make tests nearly each time I make a change. Waiting more than > a minute for this is long without clear added benefits to me. The > semantic of expect which is to check the events are there and move on > as soon as reached is precisely to minimise the time to run the > tests. The "without clear added benefit" is because the kind of event > received is logged too and checked by the regression tests and not > just the number of them. Please don't forget this is intended to be infrastructure for additional tests related to links. On the evidence to date, there is currently an immense risk of inappropriate, excessive, reporting in those cases. The basic tests have 27 instances of the 'expect' directive, hence a 54-second 'penalty'.
Last but not least those testing.c tests are > very very minimal compared to the amount of checking done by the > python tests which are way more thorough. True, but there is no link testing there that I can see. I'd expected to work with 'basic' link tests. Otherwise the python testing mechanism would need to be patched as well/instead. > running test 8 > *** ../tests/result/8 2004-10-15 09:25:29.000000000 +0200 > --- result.8 2005-09-02 17:54:58.496732932 +0200 > *************** > *** 5,12 **** > mkdir /tmp/test_gamin > 1: /tmp/test_gamin Created: NULL > mkfile /tmp/test_gamin/foo > ! 1: foo Created: NULL > rmfile /tmp/test_gamin/foo > ! 1: foo Deleted: NULL > disconnected > rmdir /tmp/test_gamin > --- 5,12 ---- > mkdir /tmp/test_gamin > 1: /tmp/test_gamin Created: NULL > mkfile /tmp/test_gamin/foo > ! expect line 8: got 0 of 1 expected events > rmfile /tmp/test_gamin/foo > ! expect line 11: got 0 of 1 expected events > disconnected > rmdir /tmp/test_gamin > running test 9 > *** ../tests/result/9 2004-10-12 22:43:05.000000000 +0200 > --- result.9 2005-09-02 17:55:08.570200863 +0200 > *************** > *** 4,12 **** > 1: /tmp/test_gamin Exists: NULL > 1: /tmp/test_gamin EndExist: NULL > mkfile /tmp/test_gamin/foo > ! 1: /tmp/test_gamin Changed: NULL > append /tmp/test_gamin/foo > rmfile /tmp/test_gamin/foo > ! 1: /tmp/test_gamin Changed: NULL > disconnected > rmdir /tmp/test_gamin > --- 4,12 ---- > 1: /tmp/test_gamin Exists: NULL > 1: /tmp/test_gamin EndExist: NULL > mkfile /tmp/test_gamin/foo > ! expect line 8: got 0 of 1 expected events > append /tmp/test_gamin/foo > rmfile /tmp/test_gamin/foo > ! expect line 14: got 0 of 1 expected events > disconnected > rmdir /tmp/test_gamin > > Reproductible, after one minute waiting for all the other previous > tests, RHEL 4 with dnotify. When I build with dnotify, no inotify, I get the same result !! Then I evaluated a bunch of wait-intervals (using usleep() instead of the loop), and found as follows: 5 intervals 0.5 ms to 15 ms pass 5 intevals 20 ms to 55 ms 8,9 fail 7 intervals 60 ms to 140 ms pass 150 ms 8,9 fail 200 ms 8,9 fail 250 ms pass 300 ms 8,9 fail Could some of those scenarios be replicated in real life (with polling)? Regards Tom _______________________________________________ Gamin-list mailing list [email protected] http://mail.gnome.org/mailman/listinfo/gamin-list
