Closed
Bug 960589
Opened 10 years ago
Closed 4 years ago
Excessive number of windows created during |make mozmill| test run of thunderbird (comm-central)
Categories
(Thunderbird :: Testing Infrastructure, defect)
Tracking
(Not tracked)
RESOLVED
WORKSFORME
People
(Reporter: ishikawa, Unassigned)
References
Details
Attachments
(1 file)
73.71 KB,
text/plain
|
Details |
Problem: I noticed that during the run of |make mozmill| test of DEBUG
BUILD of thunderbird (comm-central), windows seem to be created, but
not destroyed immediately, thus leading to a very large number of live
windows during a particular test.
++DOMWINDOW == 279 ..
(This was under 64-bit Debian GNU/Linux. I have seen larger number
over 300 under 32-bit Debian GNU/Linux. I am not sure of the
difference, but maybe there are timing differences, etc.)
The largest number 279 under 64-bit linux was
observed after the following test was run.
TEST-START | /REF-COMM-CENTRAL/comm-central/mail/test/mozmill/session-store/test-session-store.js | test_restore_single_3pane_persistence_again
I think the windows were created in large numbers before.
I recall there was a post in devel-platform about excessive # of
windows created during other tests, and the eventual culling of unused
left-over windows (they are simply left around since tests did not remove them.)
I am not sure if 279 or 300+ counts as excessive, but thought I would
report this anyway.
I am quoting the relevant log as attachment.
BTW, the particular test does not seem to close all the windows.
(As a matter of fact, it seems to start with the window # of 60, and
closing the windows down to this initial 60,
suggesting that a previous test somewhere created that many windows
and didn't bother to close them???)
Not sure if the warining/error (?) in the last part of the excerpted log
> "jum.assert", "comment": "saved state is bad so state object should be null", "value": true}
Step Pass: {"function": "jum.assert", "comment": "file should not exist", "value": true}
>
has anything to do with the excessive # of windows. I don't think so.
But, to me, that this warning/error does not seem to trip the test is more worrisome, though.
TIA
Reporter | ||
Comment 1•8 years ago
|
||
I still see some excessive number of windows created during a single test. From an output of a log summarizer. ======================================== Used DOMWINDOW ======================================== 1 ++DOMWINDOW == 362 1 ++DOMWINDOW == 361 1 ++DOMWINDOW == 360 1 ++DOMWINDOW == 359 1 ++DOMWINDOW == 358 1 ++DOMWINDOW == 357 1 ++DOMWINDOW == 356 1 ++DOMWINDOW == 355 1 ++DOMWINDOW == 354 1 ++DOMWINDOW == 353 1 ++DOMWINDOW == 352 1 ++DOMWINDOW == 351 1 ++DOMWINDOW == 350 1 ++DOMWINDOW == 349 1 ++DOMWINDOW == 348 1 ++DOMWINDOW == 347 1 ++DOMWINDOW == 346 1 ++DOMWINDOW == 345 1 ++DOMWINDOW == 344 1 ++DOMWINDOW == 343 ======================================== This large number of windows are created by the following test in C-C TB. TEST-START | /NREF-COMM-CENTRAL/comm-central/mail/test/mozmill/session-store/test-session-store.js | test_clean_shutdown_session_persistence_simple I think what happens is C-C TB creates some windows, but left them open (and possibly forgets to close?) TIA
Reporter | ||
Comment 2•8 years ago
|
||
A few permanent errors on my local TEST PC have been eliminated and exposed more issues in hitherto un-executed tests. There is an worst offender than the one in comment 1. From the summary file produced by home-brew script from the log: ======================================== Used DOMWINDOW ======================================== 1 ++DOMWINDOW == 751 1 ++DOMWINDOW == 750 1 ++DOMWINDOW == 749 1 ++DOMWINDOW == 748 1 ++DOMWINDOW == 747 1 ++DOMWINDOW == 746 1 ++DOMWINDOW == 745 1 ++DOMWINDOW == 744 1 ++DOMWINDOW == 743 1 ++DOMWINDOW == 742 1 ++DOMWINDOW == 741 1 ++DOMWINDOW == 740 1 ++DOMWINDOW == 739 1 ++DOMWINDOW == 738 1 ++DOMWINDOW == 737 1 ++DOMWINDOW == 736 1 ++DOMWINDOW == 735 1 ++DOMWINDOW == 734 1 ++DOMWINDOW == 733 1 ++DOMWINDOW == 732 ======================================== Summary ======================================== This many number of windows were created by the following test. TEST-START | /NREF-COMM-CENTRAL/comm-central/mail/test/mozmill/composition/test-multipart-related.js | test_basic_multipart_related I think 700 clearly is something we should be able to reduce. TIA
Reporter | ||
Comment 3•7 years ago
|
||
Still more windows created by TEST-START | /NREF-COMM-CENTRAL/comm-central/mail/test/mozmill/composition/test-forward-rfc822-attach.js | test_forwarding_long_html_line_as_attachment I think the function test_forwarding_long_html_line_as_attachment itself only adds a few windows, but the test functions WITHIN the file, test-forward-rfc822-attach.js, forgets to close the windows left and right, and we end up with this large number of DOMWINDOWS. ======================================== Used DOMWINDOW ======================================== 1 ++DOMWINDOW == 736 1 ++DOMWINDOW == 735 1 ++DOMWINDOW == 734 1 ++DOMWINDOW == 733 1 ++DOMWINDOW == 732 1 ++DOMWINDOW == 731 1 ++DOMWINDOW == 730 1 ++DOMWINDOW == 729 2 ++DOMWINDOW == 728 2 ++DOMWINDOW == 727 2 ++DOMWINDOW == 726 2 ++DOMWINDOW == 725 3 ++DOMWINDOW == 724 3 ++DOMWINDOW == 723 3 ++DOMWINDOW == 722 3 ++DOMWINDOW == 721 3 ++DOMWINDOW == 720 4 ++DOMWINDOW == 719 5 ++DOMWINDOW == 718 5 ++DOMWINDOW == 717 ======================================== Summary
I couldn't see any leftover windows created by these particular tests, when running them interactively. I also didn't see so large DOMWINDOW numbers. However, I do not think the number means there are so many top-level windows opened. Those seem to indicate some internal windows (we have inner and outer windows nowadays), representing frames or other partial window blocks. E.g. you can have a test opening a single compose window and the DOMWINDOW numbers still count to e.g. 30. So I do not see how to find problematic tests just by looking at the DOMWINDOW numbers.
Comment 5•4 years ago
|
||
mozmill is gone.
Status: NEW → RESOLVED
Closed: 4 years ago
Resolution: --- → WORKSFORME
You need to log in
before you can comment on or make changes to this bug.
Description
•