[PATCH] Use correct semantic DeclContext for local declarations of variables and functions with linkage

[Richard Smith]

On Wed, Aug 28, 2013 at 4:10 PM, Richard Smith <richard at metafoo.co.uk>wrote:

> On Wed, Aug 28, 2013 at 3:47 PM, Eli Friedman <eli.friedman at gmail.com>wrote:
>
>> The concept generally seems fine.
>>
>> The setLocalExternDecl thing seems a bit fiddly, since I'm pretty sure it
>> can be derived from information we already keep around.  Did you consider
>> computing that information on demand?
>>
>
> Yes, that's certainly possible. It's convenient for this to be part of the
> IdentifierNamespace, and I don't want to complicate the hot path through
> name lookup with this special case any more than necessary, but I don't
> have any benchmarks to justify doing it this way.
>
>
>> Do we compute the correct DeclContext for a friend in a local class which
>> is a redeclaration of a local extern decl?  Might be nice to test.
>>
>
> Yes, we do; I'll add a test for that.
>

This turns out to be hard (impossible?) to observe, since a local friend
must be a redeclaration of an entity in the innermost surrounding block
scope, but I've added a test for this interaction in general.


> It looks like some changes slipped in which should be committed separately
>> (around where you added the reference to 6.2.7p2).
>>
>
> I separately committed the functional part of those changes, but I'll take
> another pass and see if there's more that I can split out here.
>

Done.


> Does you patch work correctly with an extern local decl inside a block?
>>  e.g. "void f() { ^{ void extern_decl(); extern_decl(); }(); }"?
>>
>
> Yes. I'll add a test for that, too.
>

Done.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20130903/9e81ae59/attachment.html>