949891 - "ASSERTION: We expect at most one direct child frame" with <input type="number">

Status: RESOLVED FIXED

(Core :: Layout: Form Controls, defect)

Description

[Jesse Ruderman]

Attachment: testcase

###!!! ASSERTION: We expect at most one direct child frame: '!mFrames.FirstChild() || !mFrames.FirstChild()->GetNextSibling()', file layout/forms/nsNumberControlFrame.cpp, line 79

Comment 1

[Jesse Ruderman]

Attachment: stack

Comment 2

[Jonathan Watt [:jwatt]]

Attachment: patch

Comment 3

[Daniel Holbert [:dholbert]]

Comment on attachment 8360325 [details] [diff] [review]
patch

r=me, with one nit:

>+++ b/layout/forms/crashtests/crashtests.list
>@@ -48,8 +48,10 @@ skip-if(B2G) load 498698-1.html # bug 83
> asserts(1) load 578604-1.html # bug 584564
> asserts(4-7) load 590302-1.xhtml # bug 584564
> load 626014.xhtml
> load 639733.xhtml
> asserts(0-1) load 669767.html
> load 682684.xhtml
> load 865602.html
> load 944198.html
>+load 949891.xhtml
>+

nit: you're adding a fully blank line (on top of the customary terminating newline) at the end of this manifest file. Probably best not to do that. (though not a huge deal; whoever appends to this file next will presumably just fill that blank line in.)

Comment 4

[Jonathan Watt [:jwatt]]

Not having a trailing blank line really irks me. When using vim I then have to type:

 Shift-G
 Shift-$
 i
 Right-arrow
 Enter

before I can append a line instead of just:

 Shift-G
 i

So I don't agree with this convention. :)

Comment 5

[Daniel Holbert [:dholbert]]

Ah, ok -- I assumed you'd just left it there by accident.

r=me either way

Comment 6

[Daniel Holbert [:dholbert]]

(Side note: emacs actually renders the conventional trailing newline -- and so, its own "seek-to-end-of-file" command (Alt-Shift->) seeks to the cursor-position following that newline, right where you should start typing to append the new test entry.

If we followed your proposed fully-blank-line-at-the-end convention, emacs users would have to take the extra step of hitting "up-arrow" before they could start appending a line. Not a huge amount of work, sure :) but just pointing out that there's some sense to the existing convention, and changing it to save steps for you would add steps for other folks.)

Comment 7

[Daniel Holbert [:dholbert]]

Also, googling "vi commands new line" turned up the "o" command ("Open a new line after current line"), which I think saves you all of the steps in your process at the beginning of comment 4.

Comment 8

[Jonathan Watt [:jwatt]]

Ah, nice, thanks! I still think it's nicer to have a blank line ready at the end, but I can get over that a lot more easily than getting over an extra three awkward keystrokes. ;)

Comment 9

[Jonathan Watt [:jwatt]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/192c2dddd68f

Comment 10

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/192c2dddd68f

Comment 11

[Ioana (away)]

https://hg.mozilla.org/mozilla-central/diff/192c2dddd68f/layout/forms/crashtests/949891.xhtml