Summary: | Make Element::attach non-virtual | ||||||||
---|---|---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Antti Koivisto <koivisto> | ||||||
Component: | DOM | Assignee: | Nobody <webkit-unassigned> | ||||||
Status: | RESOLVED FIXED | ||||||||
Severity: | Normal | CC: | kling | ||||||
Priority: | P2 | ||||||||
Version: | 528+ (Nightly build) | ||||||||
Hardware: | Unspecified | ||||||||
OS: | Unspecified | ||||||||
Attachments: |
|
Description
Antti Koivisto
2013-08-18 09:06:59 PDT
Created attachment 209039 [details]
le patch
Comment on attachment 209039 [details]
le patch
r=me
Created attachment 209040 [details]
fix lists
Two thoughts about this that we already discussed on IRC, crystallizing here: 1. We should make the has-custom-style-resolve-callbacks bit part of the node creation flag-mask. 2. It might be good to always call the base class willAttachRenderers()/didAttachRenderers() for consistency. As long as "useless" base class implementations are inline, there will be zero overhead. (In reply to comment #4) > Two thoughts about this that we already discussed on IRC, crystallizing here: > > 1. We should make the has-custom-style-resolve-callbacks bit part of the node creation flag-mask. This is a good idea. > 2. It might be good to always call the base class willAttachRenderers()/didAttachRenderers() for consistency. As long as "useless" base class implementations are inline, there will be zero overhead. I'm not so sure about this. The call is unnecessary in most cases and calling no-op functions for "consistency" feels strange. willAttachRenderers() and pals are currently private so calling them unnecessarily will fail. Easy to test. private in Element I mean. |