Nick, thanks for cc'ing me; I'd ignored the original email because I hadn't realized it was a typo-correction patch.
Overall the patch looks good! My only little nit is that I'd prefer to keep the two ObjC-specific flags in CorrectionCandidateCallback together, though I understand why you placed IsAddressOfOperand where you did. On Mon, Jun 23, 2014 at 10:21 AM, Nick Lewycky <[email protected]> wrote: > Ping! +cc Kaelyn for typo correction review? > > On 11 June 2014 07:46, Nick Lewycky <[email protected]> wrote: > >> The attached patch fixes two bugs in one. We were correcting >> address-of-member to use the non-qualified form when it was in scope (ie., >> &Foo::typo became &corrected when we were inside Foo::method). However, >> this changes the expression from pointer to member to pointer to T (see >> [expr.unary.op]/3). This is fixed by propagating isAddressOfMember through >> the correction callback, though we now get an extra leading :: for no good >> reason. That is a separate bug, at least. >> >> The crash is because we failed to set a naming class when doing to name >> lookup of the member and finding it in a different class. Given >> &ValidButWrongOne::member we would start typo correction on "member" and >> find &ValidAndRightOne::member, we've got to mark down that the naming >> class is now ValidAndRightOne. >> >> Patch attached, please review! >> >> Nick >> >> >
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
