erichkeane added inline comments.
================
Comment at: clang/lib/CodeGen/CGExpr.cpp:200
+ E->IgnoreParenNoopCasts(getContext()))) {
+ if (CondOp->getObjectKind() == OK_BitField)
+ return EmitIgnoredConditionalOperator(CondOp);
----------------
efriedma wrote:
> Is there some reason we need to special-case bitfields here?
I don't think it ends up being a problem to support the 'simple' case (or
perhaps even the matrix/vector cases?), but I was not really motivated to write
test cases for all of those cases (nor write error handling, since that is
already so well handled).
================
Comment at: clang/lib/CodeGen/CGExpr.cpp:4635
+ eval.begin(CGF);
+ Info.LHS = EmitLValueOrThrowExpression(CGF, E->getTrueExpr());
+ eval.end(CGF);
----------------
efriedma wrote:
> If we're calling this from EmitIgnoredConditionalOperator, should we be
> calling back into EmitIgnoredExpr, instead of using
> EmitLValueOrThrowExpression? Consider, e.g. `cond ? s1.field1 = val1 : cond2
> ? s1.field2 = val2 : s1.field3 = val3;`.
First, this is the 'generic function' that is called by the normal
ConditionalOperator handling as well, so there would be a bit of branching/etc
perhaps to make this call 'ignored'.
That said, I'm not really sure what the implications of that would be. I don't
really see harm in the example you gave though, so I can put a patch together
to do that.
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D123680/new/
https://reviews.llvm.org/D123680
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
