641598 - Regression from Firefox 3.6: onblur event handler fires after calling focus()

Status: NEW

(Core :: DOM: Events, defect)

Description

[Philipp Wagner [:imphil]]

Attachment: testcase

I experienced this problem on http://www.farnell.de and reduced it to the attached testcase.

STR:
1. Load the testcase in Firefox 4 (I have here a nightly from 2011-03-14)
2. Move your mouse directly from the address bar to the input box and click in it

Expected behavior:
3a. The text inside the input should go away and the cursor blink inside the textbox.

Actual behavior:
3b. The gray text does not go away, the input field is not enabled. Entering text is not possible.


Now, to make this a bit more interesting, move your mouse around on the page, e.g. over the "button" for a couple seconds and then click in the input box again. Now the input receives focus and text input is possible.
Unfortunately, in this testcase I didn't find a 100% reproducible way to trigger to trigger the behavior that the input works again. 
On the original site however, it is more easily reproduced:

1. load http://farnell.de
2. from the address bar, go directly to the search input and click there: input does not become enabled.
3. mouseover the "Suchen" button
4. Now click into the search input: now it becomes active and it works as expected.


Firefox 3.6 shows the expected behavior, as does Konqueror 4.6.

Comment 1

[Philipp Wagner [:imphil]]

Regression window (down to a single day)

OK:
http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2010/10/2010-10-09-03-mozilla-central/firefox-4.0b8pre.en-US.linux-x86_64.tar.bz2
Built from http://hg.mozilla.org/mozilla-central/rev/d45c87e58110

NOK:
http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2010/10/2010-10-10-03-mozilla-central/firefox-4.0b8pre.en-US.linux-x86_64.tar.bz2
Built from http://hg.mozilla.org/mozilla-central/rev/26c47ba8064f

Comment 2

[Timothy Nikkel (:tnikkel)]

That corresponds to http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=d45c87e58110&tochange=26c47ba8064f

Comment 3

[Boris Zbarsky [:bzbarsky]]

Perhaps from bug 591815?

Comment 4

[Timothy Nikkel (:tnikkel)]

(In reply to comment #3)
> Perhaps from bug 591815?

Yep, verified by undoing that patch.

Comment 5

[Timothy Nikkel (:tnikkel)]

The NS_MOUSE_BUTTON_DOWN case of nsEventStateManager::PostHandleEvent doesn't seem to sanely handle the case where mCurrentTarget, aTargetFrame, and mCurrentTargetContent are all null.

Comment 6

[Timothy Nikkel (:tnikkel)]

Specifically the case where nsEventStatus_eConsumeNoDefault != *aStatus.

Comment 7

[Timothy Nikkel (:tnikkel)]

Bug 591815 was only about the prevent default case, so maybe for this case we should make the processing conditional on having a target frame like it was before bug 591815?

Comment 8

[Timothy Nikkel (:tnikkel)]

Attachment: diff -w

Should we just do this?

Comment 9

[Timothy Nikkel (:tnikkel)]

Attachment: patch

Comment 10

[Olli Pettay [:smaug][[e-mail]]]

-w patch might make reviewing easier.

Comment 11

[Timothy Nikkel (:tnikkel)]

There is a -w patch: https://bugzilla.mozilla.org/attachment.cgi?id=520059

Comment 12

[Olli Pettay [:smaug][[e-mail]]]

Oops, sorry, I somehow missed that comment.

Comment 13

[Olli Pettay [:smaug][[e-mail]]]

Comment on attachment 520060 [details] [diff] [review]
patch

We really need some tests here.

r=me if you add tests.

Comment 14

[Timothy Nikkel (:tnikkel)]

Attachment: patch for checkin

Comment 15

[(no longer active)]

Timothy, have you posted this to the try server?

Comment 16

[Timothy Nikkel (:tnikkel)]

I have not.

Comment 17

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

Looks like this causes a M1 timeout: <http://tinderbox.mozilla.org/showlog.cgi?log=MozillaTry/1303400192.1303403351.30016.gz>

Comment 18

[Timothy Nikkel (:tnikkel)]

Thanks for sending that to try server.

Comment 19

[Olli Pettay [:smaug][[e-mail]]]

Timothy, do you time to work on this?

Comment 20

[Timothy Nikkel (:tnikkel)]

Not at the moment, I was planning to come back to this, but if you want to pick it up feel free to do so.

Comment 21

[Timothy Nikkel (:tnikkel)]

Attachment: unbitrotted

Still fails
ERROR TEST-UNEXPECTED-FAIL | /tests/content/events/test/test_bug534833.html | Test timed out.

Comment 22

[Timothy Nikkel (:tnikkel)]

Attachment: updated testcase to keep it working

Comment 23

[Timothy Nikkel (:tnikkel)]

Attachment: reduced testcase

Comment 24

[Timothy Nikkel (:tnikkel)]

This is now working. I'm not sure what fixed it, I think it's only working by accident because the fix ranges I get on mac and windows differ

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=582ea6610e8aa4120ff223167ade012381bf4dae&tochange=3d9044e7bb3f7e3e6d261cdefaea6e647068d3e5

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=177ac92fb734b80f07c04710ec70f0b89a073351&tochange=206721f8064a145ddb526eb9c50285fd274e7d87

And I've double or triple checked those ranges. So this could come back or similar but different testcases could fail.