363675 - reflow branch landing changed width behavior of abs/fix pos iframes

Status: RESOLVED DUPLICATE of bug 412679

(Core :: Layout: Positioned, defect)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

The reflow branch landing changed the width behavior of absolute and fixed positioned iframes -- at least those with left:0 and right:0, so that they use their intrinsic size (i.e., follow the rules for replaced elements) rather than stretching (following the rules for non-replaced elements).  This caused bug 363154.

Testcases to follow, then analysis of behavior of other browsers.

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: position: absolute; all offsets 0

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: position: fixed; all offsets 0

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

This is whether the following browsers expand the positioned elements to fill their containing block.  I have not found any differences between the position:absolute and position:fixed cases.

Firefox 2: expands both height and width
WinIE7: expands neither
Konqueror: expands neither
Opera: expands only height
Firefox trunk (post reflow branch): expands only height

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: position: absolute; all offsets 0 (quirks mode)

Comment 5

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: position: fixed; all offsets 0 (quirks mode)

Comment 6

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

The quirks mode results are the same.

So I conclude we should probably keep the changed behavior, and make the same change for height, since I think the new behavior agrees both with the spec and with other browsers.

Comment 7

[Boris Zbarsky [:bzbarsky]]

Yeah, agreed.

Comment 8

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

(That said, there are advantages to the old behavior.  The 300*150 stuff is completely arbitrary and it would be nice to have it show up in fewer cases.  Maybe I should try to find a way to get back to it.  But if we do, we should probably try to change the spec.)

Comment 9

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I'd have thought implementing ComputeAutoSize on nsLeafFrame would fix this; turns out it doesn't.

Comment 10

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: I'd have thought this would have worked, but it didn't

Comment 11

[Wladimir Palant]

David, while sticking to the spec is generally a good thing - is there any other way to stretch a frame vertically to window size while keeping offsets on both sides? position:absolute;left:0;margin:10px;width:100%; stretches the document and introduces scrolling so it cannot be a replacement for position:absolute;left:10px;right:10px;. Unless I am missing an obvious way to do this I would rather vote for keeping the behavior from Firefox 2 and changing the spec.

Comment 12

[Boris Zbarsky [:bzbarsky]]

The issue with changing the spec is that it involves convincing all the other browsers to change their behavior (which currently complies with the spec).  So it's not so much a matter of "we should follow the spec" as "we should have compat with the way every single other UA does it"....

I agree that the use case you describe is useful, and there should be a way to achieve it (say a -moz property that says to not use replaced element sizing or whatever).

Comment 13

[Anne (:annevk)]

The issue from comment 11 can probably be addressed by using the calc() property from CSS3. I agree though that the behavior from Firefox 2 was much more what you would expect. I suppose the main problem with that one is convincing IE to do the same, as you say...

Comment 14

[Martijn Wargers (dead)]

I don't think this works at all in IE7, see: http://wargers.org/mozilla/abspos.htm
That shows it's not even working for a <div>.
So it's probably not useful to look at what IE is doing.

Comment 15

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: baseline testcase with div: position: absolute; all offsets 0

Comment 16

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: baseline testcase with div: position: fixed; all offsets 0

Comment 17

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: baseline testcase with div: position: absolute; all offsets 0 (quirks mode)

Comment 18

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: baseline testcase with div: position: fixed; all offsets 0 (quirks mode)

Comment 19

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

WinIE7 gets the standards mode baseline testcases correct but not the quirks mode ones.  The other browsers tested get all the baseline testcases correct.

Comment 20

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Looks like Boris has a patch for this n bug 412679.

Comment 21

[Boris Zbarsky [:bzbarsky]]

Hmm... And it's pretty identical to the one you had here.  I wonder why it didn't work for you?

Anyway, I'm happy with taking the assertion changes here that explain things in a little more detail than the assertions in my patch.  I do think we should make the nsFrame.h comment change in my patch...

Comment 22

[Boris Zbarsky [:bzbarsky]]

So should we mark this duplicate of bug 412679?