On Fri Dec 6 12:22:50 2024 +0000, Jacek Caban wrote:
Please avoid casts like this. The usual approach for similar node types is to store a member like `nsIDOMAttr *dom_attr` in `HTMLDOMAttribute`. This change would also use better patch splitting. For instance, you could first introduce the `dom_attr` member and incrementally use it in places like this. Once those are updated, you could convert the object into a fully-fledged MSHTML node. Also, those changes are not currently covered by tests. We could probably run `test_attr` from `dom.c` in IE11 or IE9 mode.
Won't splitting it introduce temporary regressions? Some of them will use gecko attributes, while the others will use mshtml attributes, and so will be out of sync?