On Jun 9, 2011, at 2:49 AM, Maciej Stachowiak wrote: > > On Jun 8, 2011, at 11:48 AM, Peter Kasting wrote: > >> On Tue, Jun 7, 2011 at 11:18 PM, Maciej Stachowiak <[email protected]> wrote: >> 1) We definitely have consensus to fix the broken non-logically-const >> accessors by making them non-const; consensus on also adding const accessors >> is less clear. >> >> There are a surprising number of places that actually do const traversals. >> Simply making all these accessors non-const will require removing a lot of >> valid const usage from the existing code. I'm really uncomfortable with >> that. >> >> 2) We like to do things incrementally and fixing bad use of const before >> adding good use of const seems like this is a logical way to split the work >> and make it less of a megapatch. >> >> Incremental fixes are absolutely the way to go. Reviewing megapatches sucks >> and it's hard to catch subtle bugs like "you changed this function to be not >> const, but there's no reason to do that". >> >> Perhaps a split that avoids removing existing, valid const usage would be to >> first change few (or no) function signatures, and simply modify caller code >> to be more consistent about type declarations. This would mean converting >> some callers from "Node*" to "const Node*" when they're doing true const >> traversals, and some the opposite direction when they're not. The goal >> would be to make eventual API changes a no-op in terms of caller effect. >> It'd be easy to make these sorts of patches arbitrarily small as well. > > How about posting a reasonable-sized patch that handles a few related, > representative methods so we can evaluate.
Now having read the rest of the thread, I think you should start by posting what you and Darin agreed to. - Maciej
_______________________________________________ webkit-dev mailing list [email protected] http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
