On Mon Aug 14 13:35:11 2023 +0000, Jacek Caban wrote:
DOM Nodes don't make use of the destructor yet, so they are special case
no-ops as temporary exceptions, because they still use their own mechanism which will change in the future. I think that this patch still needs more work on splitting. I'd be also more interested in how nodes will look like than in module-wide change skipping interesting parts. In fact, I'd expect that it would be nicer to convert nodes first and use that to convert other objects more incrementally. If you do it the other way around, you limit yourself about how you may split nodes later. The first patch for nodes could be as simple as moving ccref to DispatchEx. As long as it's still initialized out of init_dispatch, it will be optional. Then you could move node's CC participant to dispex.c with a new helper than will ultimately replace `dispex_query_interface`. Once you'd have nodes moved like that, you could convert other objects one by one. It's fine to group trivial ones, but not too much. BTW, I think that working with dispex vtbl would be nicer if we converted them to designated initializers. Also, as ccref becomes a dispex internal thing, I'd suggest to drop 'cc' from addref/release helper names. This MR is long anyway, please drop this patch from this MR.
Oh I didn't realize it was fine to temporarily use a different helper for `dispex_query_interface`, ultimately ending up back to it I suppose once all objects are converted. In that case yeah, I could convert nodes first.