I filed https://github.com/mozilla/servo/issues/1676 a few days ago but
there hasn't been any discussion yet; I think it's important to sort out
this problem now so we don't have to do a lot of backtracking when we
get around to implementing the COW DOM.
In brief, we have various methods in layout that cast a node to its
concrete type and fetch a data member (such as HTMLImageElement's
|image|, which is a Url that's set in content when the src attribute
changes). As of right now, this particular operation is racy, since we
don't have any calls to Node::wait_until_safe_to_modify_dom in place
around the writing code in script to guarantee exclusion.
This brings up the larger point that we're doing this same thing in
various places, and we need to sort out a strategy. In the COW DOM
world, we would either need to implement COW for certain Node-derived
members or declare that layout always fetches the attribute value, which
will be set up for COW. This latter strategy still means that properties
such as CharacterData's content will need to be COW, since they're not
attribute-based.
For the short term I'm interested in making layout code always go for
the attributes, because relying on sprinkling
wait_until_safe_to_modify_dom all over content code is easy to get
wrong. Thoughts?
Cheers,
Josh
_______________________________________________
dev-servo mailing list
dev-servo@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-servo
