stylesheet loading

[David Hyatt]

On Tuesday, June 3, 2003, at 02:40 AM, Dirk Mueller wrote:

> Hi,
>
> while trying to fix some of the stylesheet loading regressions that 
> occurred
> after the FOUC merge, I wondered about this diff chunk in
> html/html_headimpl.cpp :
>
> -    if(type.contains("text/css") || rel.contains("stylesheet")) {
> +    // This was buggy and would incorrectly match <link 
> rel="alternate">,
>      // which has a different specified meaning. -dwh
> +    if(m_disabledState != 2 && (type.contains("text/css") || rel ==
>      "stylesheet" || (rel.contains("alternate") && 
> rel.contains("stylesheet"))))
>
>
> I can't see how the previous code was able to match <link 
> rel="alternate">.
>

We are operating off of KDE 3.0.2 remember.  In that tree, the code 
looked like this.

if(type.contains("text/css") || rel == "stylesheet" || 
rel.contains("alternate"))

It looks like you later fixed it.  The above would have incorrectly 
matched <link rel="alternate"> (which is used for RSS files).


> My Bloatzilla (1.2.1) behaves like Konqueror previously did, i.e.
>
> <link rel="stupid stylesheet">
>
> will get loaded.
>

We could certainly relax the restriction, especially if Mozilla allows 
it.

> Also, this m_disabledState handling seems quite broken to me. it 
> stores a
> "modifiedbyJScript" state which is nowhere used.. Anyway, disabling is 
> just
> a boolean and you cannot make a triple state from this information 
> alone.
> I'm going to remove this code again as it breaks dynamic disabling of
> stylesheets.
>

The state is very much used in order to properly increment and 
decrement the pending sheet count when stylesheets get enabled/disabled 
dynamically prior to the page having finished loading (or even the 
stylesheets having finished loading).  See parseAttribute (the 
ATTR_DISABLED code).

Can you give me an example of a site that's broken?  That will tell me 
whether there's a merge problem or a bug in the actual code in Safari.

Thanks!
dave
(hyatt at apple.com)