67690 - aReflowState.mComputedWidth/Height incorrectly computed when absolutely positioned

Status: VERIFIED FIXED

(Core :: Layout: Form Controls, defect)

Description

[jsp]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; m18) Gecko/20010125
BuildID:    2001012504

Checkbox and radio button controls are larger when positioned absolutely than
when they are not so positioned. Size and position should not be related.

Reproducible: Always
Steps to Reproduce:
Create a form with an absolutely positioned checkbox or radio button and another
that is not absolutely positioned. See the attachment for an example.

Actual Results:  The positioned control is +/- 25% larger than the unpositioned
control.

Expected Results:  Both controls should be the same size.

The same behavior occurs on Macintosh build 2001020208.

Comment 1

[jsp]

Attachment: Test case.

Comment 2

[James Lariviere]

WOW, Wild.  Confirming new.

Comment 3

[Boris Zbarsky [:bzbarsky]]

The label positioning may actually be right...  The top of the controls lines up
with the top of the text if you ignore ascenders.

But the sizing issue is weird.

Comment 4

[jsp]

Attachment: Screen shot of incorrect test case results.

Comment 5

[jsp]

I think the label positioning is correct; it's the positioning of the oversized
radio button that's suspect. See the new screen shot attachment.

Comment 6

[rods (gone)]

Just as it says, I will attach an example using div. Checkboxes and radiobutton 
use nsLeafFrame:Reflow which eventually gets the size from 
aReflowState.mComputedWidth and it is there that it is already calculated 
incorrectly.

Comment 7

[rods (gone)]

Attachment: testcase with divs

Comment 8

[Marc Attinasi]

Interestingly, relative positioning is OK. I'm investigating this now...

Comment 9

[Marc Attinasi]

Problem is actually with box-sizing in absolute positioned items. I have a patch 
for it now, will attach. Basically, box-sizing was being ignored in 
nsHTMLReflowState::InitAbsoluteConstraints.

Comment 10

[Marc Attinasi]

Attachment: Proposed patch: honor box-sizing value

Comment 11

[Chris Waterson]

Should make two functions (one for width, one for height) that incorporate this 
with the code which respects the min/max width/height? (Possibly inline.) Also, 
how about ``apply padding and borders unless unconstrained'' instead of ``MJA - 
from CompueBlockBoxData''. :-)

Comment 12

[buster]

yep, what chris said

Comment 13

[Marc Attinasi]

OK - thanks. I'll consolidate the code, and there are a few other places that
can use it too.

Comment 14

[Marc Attinasi]

Attachment: Patch: coding like a responsible adult instead of a giddy adolescent

Comment 15

[Chris Waterson]

looks good to me. run them regression tests, and r=waterson

Comment 16

[Marc Attinasi]

Regerssion tests are OK - one meaningful difference was found,
 * file:///s:/mozilla/layout/html/tests/block/bugs/55874.html
The rendering is different but more correct. I'm also checking in the testcase
for this bug for future regerssion testing.

Comment 17

[Marc Attinasi]

Marking Mozilla 0.9 since I have a reviewed, approved patch in hand.

Comment 18

[Marc Attinasi]

Milestone shift --> Moz 0.8.1

Comment 19

[David Hyatt]

sr=hyatt

Comment 20

[Marc Attinasi]

Fix checked in: nsHTMLReflowState.h/cpp

Comment 21

[jerone]

Verified build 20010724& 20010730trunk os:mac8.6/win98/linux7