Bug 4959 – inline style not terminated properly after <form>

• Status: RESOLVED
• Resolution: FIXED
• Component: Layout and Rendering

[Antti Koivisto]

In http://etuovi.com/ the content of the box under headline "pikahaku" is missing. This is because in a 
following kind of construct

<span style="display:none"><form></span>this stuff should be visible but is not

the style specified in inline element does not terminate on close tag, but continues to be a applied to 
following elements

Works on IE/Gecko

[Antti Koivisto]

Created attachment 3882 [details]
test case

[Antti Koivisto]

Created attachment 3883 [details]
add span as a residual style tag and form as an affected tag

add <span> as residual style tag
add <form> as tag affected by residual style

Actually the second one is enough to fix the page refferred in the bug report
(it uses <font>) but the first one is needed to fix the linked test case. The
first one causes changes in the results of some regression tests (but those may
be harmless as the cases are still visually correct):
editing/deleting/delete-image-003	
editing/undo/redo-typing-001	
editing/undo/undo-typing-001
fast/block/float/031

[Darin Adler]

Comment on attachment 3883 [details]
add span as a residual style tag and form as an affected tag

I'm pretty sure this is not exactly right, because there are many other tags
besides spans that can result in an inline element, for example any custom
non-HTML tag will create an element with the same behavior as the one created
by a span.

But Dave Hyatt is the expert in this area, so waiting for input from him.

[Antti Koivisto]

Agreed, it is probably not right. However the whole mechanism of special casing some html tags for 
different treatment looks rather suspicious to me. At least based on this test case <span> and <b> 
shouln't be treated any differently, as they are now. 

[Dave Hyatt]

Comment on attachment 3883 [details]
add span as a residual style tag and form as an affected tag

r=me, but performance tests will need to be run for this one, since it will
probably result in the creation of more elements in our page load tests.

[Darin Adler]

I asked Tim to land a couple patches that have been here for a long time, and he agreed to do it.

[Timothy Hatcher]

Performance tests not affected. Landed in r12508.

[Dave Hyatt]

fast/block/float/031.html fails right now with actual visual differences in rendering.

Given that this is one of the tests listed as being supposedly unaffected visually, I'm highly suspicious. :)

[Dave Hyatt]

Yes, <span> cannot be reopened due to residual style.  I verified that this fix broke fast/block/float/031.html, and that the rendering of fast/block/float/031.html is correct without this fix.

I am going to back out only the <span> portion of this bug fix, since as the reporter described, it was not needed for this bug anyway.

(The snippet in fast/block/float/031.html with its crazy mismatched span/a nonsense came from a real-world Web site.)

[Dave Hyatt]

I backed out the <span> portion of this fix.