DHowett-MSFT planned changes to this revision.
DHowett-MSFT marked an inline comment as done.
DHowett-MSFT added inline comments.
================
Comment at: lib/Sema/SemaExpr.cpp:7749
+ // id (or strictly compatible object type) -> T^
+ if (getLangOpts().ObjC1 &&
RHSType->isBlockCompatibleObjCPointerType(Context)) {
Kind = CK_AnyPointerToBlockPointerCast;
----------------
theraven wrote:
> Do we want to allow implicit casts for all object types to block types for
> assignment, or only for null pointers? We definitely want to allow `nil` to
> be assigned to a block type, but I would lean slightly to requiring an
> implicit cast.
>
> Ideally, I think we'd allow this but warn, because casting from an arbitrary
> ObjC type to a block incorrectly can cause exciting security vulnerabilities
> if it's done incorrectly and we should encourage people to check these casts
> (`nil` is always safe though - as long as somewhere else checks the
> nullability attributes).
I don't actually have a compelling use case for widening assignability here.
I'm dropping this part of the patch.
I do think it should be valid, perhaps with a warning. It feels incorrect for
there to be valid comparison cases that are not valid assignment cases (here),
but that position doesn't hold up under scrutiny.
Repository:
rC Clang
https://reviews.llvm.org/D44580
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
