322066 - The script generated contents are lost by replacing frames

Status: RESOLVED WONTFIX

(Core :: XBL, defect)

Description

[Masayuki Nakano [:masayuki] (he/him)(JST, +0900)]

If a XBL component has generated contents by script (e.g., the access key of checkbox of toolkit.), these contents are gone when the frame is replaced by chenged style (e.g., opacity, position).

Comment 1

[Masayuki Nakano [:masayuki] (he/him)(JST, +0900)]

Attachment: testcase

When you hover the mouse on the checkbox, the accesskey is gone from its label.

Comment 2

[Boris Zbarsky [:bzbarsky]]

So is the problem that the binding is recreated when the style changes in this case?  Or something else?

Comment 3

[Masayuki Nakano [:masayuki] (he/him)(JST, +0900)]

I don't know what way is best approach. Because I don't understand current frame replacing mechanism...

Comment 4

[Boris Zbarsky [:bzbarsky]]

Well, I don't know where that accesskey display comes from.  Where does it come from?  I see nothing in the XBL that would do it; am I just missing it?

Comment 5

[Masayuki Nakano [:masayuki] (he/him)(JST, +0900)]

http://lxr.mozilla.org/mozilla/source/toolkit/content/widgets/text.xml#93
> 93       <method name="formatAccessKey">
That is generated here.

Comment 6

[Boris Zbarsky [:bzbarsky]]

So is the constructor being called on the frame reconstruct here?  Or no?

Comment 7

[Masayuki Nakano [:masayuki] (he/him)(JST, +0900)]

the constructor isn't called.

Comment 8

[Boris Zbarsky [:bzbarsky]]

Interesting.  DOM inspector doesn't show the text node in question in the anonymous tree even for the case when the accesskey is being shown.  It'd really help to have a minimal-ish testcase here (as little XBL as possible).  The "text" binding is huge.

Comment 9

[Martijn Wargers (dead)]

Attachment: testcase2

I failed miserably making a minimal testcase, but I think this bug is basically the same as bug 255966.

The testcase I made is almost the same as the testcase in bug 255966, but instead of toggling display:none/display:block, I use here position:relative.

Comment 10

[Masayuki Nakano [:masayuki] (he/him)(JST, +0900)]

Umm... I look like it is same bug.
But bug 255966 is worked by nobody in this 1.5 years...

If it's difficult, we need another approach for bug 306763.
Because it's accessibility issue.

Do someone have a plan to fix this bug and bug 255966?

Comment 11

[Boris Zbarsky [:bzbarsky]]

> Do someone have a plan to fix this bug and bug 255966?

In what timeframe?  For 1.8.x, no.  For 1.9 we're hoping to bring sanity to XBL insertion points, which would likely help (or at least make this easier to debug).

Comment 12

[Boris Zbarsky [:bzbarsky]]

XBL is now disabled in Firefox (Bug 1583314) and is in the process of being removed from Gecko (Bug 1566221), so closing bugs requesting changes to its implementation as wontfix.