Community
Participate
Working Groups
Verifying performance results for I20090317-1029, I see that there's a regression on 'Draw on a window' on Win/XP machine. The test results is currently grayed in the fingerprint with the following comment: 'Regression due to a issue on the Releng test machine' If I remember well, this comment was added to explain a regression occurring on a 3.4 test machine but was incorrectly released in HEAD stream. So, it should be removed now or changed to a new reason (e.g. looking at history, surely a repeatability problem).
It's only repeatable on the releng test machine. When I run the test locally I get consistent results. Furthermore, the results on the releng test machine don't make sense to me. The mean result there is 14.39 seconds, but on my XP box (a much slower machine) the Elapsed time is 1.12-1.15 seconds. How come this test runs so slowly on the test machine? How are the tests run on the releng machine? What kind of video card does that machine have? Is there a screen saver or something else running that is preventing windows from drawing properly?
(In reply to comment #1) > It's only repeatable on the releng test machine. When I run the test locally I > get consistent results. > Furthermore, the results on the releng test machine don't make sense to me. The > mean result there is 14.39 seconds, but on my XP box (a much slower machine) > the Elapsed time is 1.12-1.15 seconds. How come this test runs so slowly on the > test machine? How are the tests run on the releng machine? What kind of video > card does that machine have? Is there a screen saver or something else running > that is preventing windows from drawing properly? > Kim, would it be possible that some background processes were running on the XP machine polluting the results?
The linux test machines are much faster also. SLED: 1.04s RHEL: 1.93s
(In reply to comment #3) > The linux test machines are much faster also. > SLED: 1.04s > RHEL: 1.93s > Kevin, can you reproduce the same differences locally (i.e. while running this test from an eclipse session)?
That is how I run the tests
ok. I hoped that when the Windows releng machine was cleaned a little while ago, this test would go back to normal, but it seems not to be the case. So, I guess we could keep the test commented and close this bug...