202138 - deleteRule does not work with @import

Status: RESOLVED WORKSFORME

(Core :: DOM: CSS Object Model, defect)

Description

[Garrett Smith]

User-Agent:       Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.3) Gecko/20030312
Build Identifier: Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.3) Gecko/20030312

rule 0 is "@import 'vegBg.css'"

vegBg.css contains a backgroung image.

deleteRule(0) deletes the rule, but the background image stays (so do other
properties).

Reproducible: Always

Steps to Reproduce:
1. try the demo.
2.
3.

Actual Results:  
I see font-style: oblique and veggie bg image.

Expected Results:  
when imported rule is deleted, we should see a visual result.

Comment 1

[Garrett Smith]

Attachment: imported css for following attachment

Comment 2

[Garrett Smith]

Attachment: background-image for veg.css

Comment 3

[Garrett Smith]

Attachment: imported stylesheet used in following example

fixed url for background-attachment to point to bugzilla's location

Comment 4

[Garrett Smith]

Attachment: Shows how @import deletion fails [THIS IS IT!!!]

Comment 5

[Boris Zbarsky [:bzbarsky]]

So.. I assume deleting the @import rule _is_ supposed to remove the child sheet?
 (The spec is not exactly very clear here....)

Comment 6

[Boris Zbarsky [:bzbarsky]]

Over to Ian, since "qawanted" is not having its desired effect.  ;)

Comment 7

[Hixie (not reading bugmail)]

Well what else could it do?

Comment 8

[Boris Zbarsky [:bzbarsky]]

It could also remove the rule from the sheet, such that when the sheet is
serialized it is no longer present, but without affecting the page's stylistic
information.

Comment 9

[Hixie (not reading bugmail)]

Seems like that would be rather inconsistent with the effect of removing any of
the other rules in the sheet, which we agree does affect the rendering.

Comment 10

[Boris Zbarsky [:bzbarsky]]

Removing @charset rules does not.  Removing @namespace rules may or may not
(it's not clear what the hell it should do exactly).

Comment 11

[Hixie (not reading bugmail)]

Those two affect the parsing, not the semantics of the document. At least,
that's my line and I'm sticking to it until there's a spec that disagrees with
me. :-P

Comment 12

[Boris Zbarsky [:bzbarsky]]

In case you care, I checked.  In our impl, _inserting_ an @namespace rule makes
it take effect, while removing one does not make it not take effect.

And it DOES affect the semantics.  Consider the following DOM2 calls:

sheet.insertRule("@namespace foo url(bar);", 0);
sheet.insertRule("foo|div { color: green }", sheet.cssRules.length);
sheet.deleteRule(0);
sheet.insertRule("foo|div { color: red }", sheet.cssRules.length);

The div in the foo namespace should be green.

So my point is, the spec does not clearly describe what should be happening here
(and I too could pull out the "nothing should change and there is no spec that
disagrees with me" crap...  ;) )

It does make sense to me that the child sheets would be removed when the @import
rule is removed, though....  

Comment 13

[Hixie (not reading bugmail)]

> sheet.insertRule("@namespace foo url(bar);", 0);
> sheet.insertRule("foo|div { color: green }", sheet.cssRules.length);
> sheet.deleteRule(0);
> sheet.insertRule("foo|div { color: red }", sheet.cssRules.length);
> 
> The div in the foo namespace should be green.

(I assume you mean "bar namespace".)

What should happen in the CSSOM with namespaces is totally undefined. I agree
with what you say should happen, but that's not changing the semantics of the
existing rules, it's changing the semantics of parsing future rules, which isn't
want I meant by "changing semantics".

The point being that once you parse the "foo|div" selector, the concept of "foo"
is lost, the selector becoming internally equivalent to "{bar}foo" -- the prefix
is no longer relevant. Upon serialising, a UA will probably want to be clever
and look at what prefixes are available, and serialise using the appropriate one
(and whine if there isn't one for that namespace), but that, again, is not a
change to the semantics.


As for this bug, I think we're agreed -- changing @import rules should affect
the actual import, live.

Comment 15

[Paul Silaghi, QA [:pauly]]

Dropping qawanted since it was satisfied in comments 7-13

Comment 16

[Garrett Smith]

FF 24 on Ubuntu.

The desired effect of the stylesheet being deleted seems to be taking effect, live.

The vegBg, declared with a higher specificity (HTML > BODY), is not shown. Instead see the red background declared with (BODY).

I'm am declaring this RESOLIVED FIXED.

User-Agent: "Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:24.0) Gecko/20100101 Firefox/24.0"