1565330 - |make mozmill| execution of newmailaccount/test-newmailacccount.js: JavaScript error: chrome://messenger/content/tabmail.xml, line 732: TypeError: aTabMode is undefined

Status: RESOLVED WORKSFORME

(Thunderbird :: Account Manager, defect)

Description

[ISHIKAWA, Chiaki]

Attachment: TEST-START, TEST-UNEXPECTED-FAIL and the relevant error lines.

This is with comm-central fetched within last 48 hours.
I am testing FULL DEBUG version of C-C TB locally under linux (64-bit compilation).
However, I believe the issue will be on other platforms as well.

When I ran |make mozmill| test, I see more than 460+ failed tests.
Possibly the major reason for the failure could be
bug 1565321, and bug 1565327.

But the problem I am reporting here is very serious.

I see the following error during the execution of |make mozmill| test suite when the test is performed
with newmailaccount/test-newmailaccount.js.

JavaScript error: chrome://messenger/content/tabmail.xml, line 732: TypeError: aTabMode is undefined

Not all of the tests in the file sees the error, but some see the errors twice.

Also, maybe because of this error, we have a timeout for each test in the file.
I see the following error.

JavaScript error: file:///NREF-COMM-CENTRAL/mozilla/comm/mail/test/mozmill/shared-modules/test-window-helpers.js, line 365: Error: Timeout while waiting for modal dialog.

In the attached file, I only extract the relevant lines from the log to show the relationship of the errors and tests.
I used the following to extract the start/end of each test, and the relevant error lines.
egrep "(JavaScript error: chrome://messenger/content/tabmail.xml, line 732: TypeError: aTabMode is undefined|^TEST-|Timeout while waiting)" log1022-mozmill.txt

TIA

Comment 1

[Jorg K (CEST = GMT+2)]

I think I've see this. Will that go away with tabmail de-XBL? Or carried over to the new JS goodness?

Comment 2

[ISHIKAWA, Chiaki]

One note: I have noticed some drawing operations never finishes before |make mozmill| moves on to the next operation in some of the tests performed by it.
I can tell because I see a frame of a new window, but before it is fully populated and complete image is visible,
it disappears and the next window frame appears and this is repeated in certain tests. (Not necessary this particular test. This particular tests do not show such flickering window frame.)

I wonder if the proper delay or wait is necessary before such drawing is completely finished and visible.
If TB wanted to access, say, a dialog and if it is not visible, it cannot be responded. Correct?

I noticed such unfinished drawing several years ago during testing which were subsequently fixed, but
certain too-quick-to-move-to-next-step behavior has crept in the last six months or so (or maybe my X11 has been upgraded and show the behavior more prominently).

Comment 3

[Magnus Melin [:mkmelin]]

Probably carries over, but who knows. Sounds like some kind of timing issue, and timings may be slightly different.

Comment 4

[Magnus Melin [:mkmelin]]

mozmill is gone, and if the problem still exists, it would look different