44434 - width and height attributes appear to be mistyped

Status: VERIFIED FIXED

(Core :: DOM: Core & HTML, defect, P2)

Description

[Dylan Schiemann]

Overview Description: JavaScript stores width and height attributes as numbers, 
but the latest nightly builds fail to take stored width and height values in 
rendering an image

Steps to Reproduce: 
1) Create an image object in JavaScript and set its width and height (note that 
setting them to strings will give an error message)
2) Create an image elements using the dom and then set the image width and 
height to the values saved in the image object
3) appendChild() fails to display an image.

Actual Results: 
Image does not appear

Expected Results: 
Image should appear with stated width and height.  This is a recent break as it 
worked as expected in the M16 release, but has ben broken for over a week in the 
nightlies.

Reproducibility: 
Always

Build Date & Platform Bug Found: 
20000629, WinNT

Additional Builds and Platforms Tested On: 
20000701, Win2k, 2000062608
Works in M16 milestone build

Additional Information: 
Testcases to follow

Comment 1

[Dylan Schiemann]

Attachment: testcase

Comment 2

[Dylan Schiemann]

Attachment: more general testcase

Comment 3

[Dylan Schiemann]

Attachment: ignore first two testcases... missed a bug

Comment 4

[Dylan Schiemann]

Attachment: workaround

Comment 5

[Johnny Stenback (:jst)]

Attachment: Proposed fix

Comment 6

[Johnny Stenback (:jst)]

I just attached a fix for this problem, the problem is that mozilla never
returned the width&height (through img.width & img.height) of images that
weren't displayed (i.e. there's no frame for the image), I added code that gets
the width&height from the attributes if there's no frame to ask. Nominating for
nsbeta2, this fix is very low risk.

Comment 7

[Jay Patel [:jay]]

Putting on [nsbeta2+] radar for beta2 fix.

Comment 8

[Johnny Stenback (:jst)]

The fix is in!

Comment 9

[Dylan Schiemann]

It works great... thanks

Comment 10

[Jim Roskind]

Dylan's exclamation above sure come close to a "verification" of the fix.
I would have done some verification if it was clearer how to interpret the "test
case" to tell if the fix is working.
Specifically: What constitutes failure in the test case? What constitutes
success in the test case?
Thanks,
Jim (trying to help) Roskind

p.s., The test case does produce a nice square, which rapidly vanishes leaving
only the text "blank" showing.  I suspect that the goal is to see a square...
but I was not sure.  It might be nicer if the square actually had some color, so
that it could be seen on the background after its outline (supplied during
progressive layout) vanished.

Comment 11

[Dylan Schiemann]

The testcase needs access to an image file to work effectively... I could not 
find a way to get Bugzilla to reference an uploaded image.  Attached is a 
better testcase as it attempts to stuff the mozilla.org logo into the box.  
Failure in older builds resulted in no rendering of the image when the width 
and height were set with the values stores in an Image() object.  A "real 
world" testcase is the navigation interface for 
http://www.westcoastinternet.net  Several of the graphics on that site display 
only in versions after the fix was introduced (fail in 20000626, work in 
20000801.  Thanks, Dylan

Comment 12

[Dylan Schiemann]

Attachment: better testcase

Comment 13

[Jim Roskind]

Verified on NT with build 2000080204 (M17 beta2 branch).
The logo gets stuffed into the box... and is visible.

Comment 14

[gerardok]

Verified on 2000-08-01-05-M17 build on Win2k.