On Tue, Jun 18, 2013 at 11:05 AM, Bill Wendling <[email protected]> wrote: > On Jun 18, 2013, at 7:29 AM, David Blaikie <[email protected]> wrote: > >> >> On Jun 18, 2013 12:24 AM, "Bill Wendling" <[email protected]> wrote: >> > >> > Author: void >> > Date: Tue Jun 18 02:22:05 2013 >> > New Revision: 184176 >> > >> > URL: http://llvm.org/viewvc/llvm-project?rev=184176&view=rev >> > Log: >> > Update to pass in pointers intead of references. >> >> Why? >> > Basically, these objects may change during compilation. This allows us to > change them when that happens.
Sorry, yes, I eventually got to your corresponding commits in LLVM & replied there too. (copying that sentiment here in case others have opinions): Personally I'd rather not see parameters change from references to pointers when the semantics aren't changing - it's just the implementation that's changing so the member pointers can be repointed. When I see pointers as parameters I have to consider whether they can be null & what the semantics are in that case. (I realize this is not a strong philosophy in the LLVM codebase in general, so I don't have much of a strong position to argue from - but it's a bit of a pity when we go backwards from where I would like to be) _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
