[PATCH] Don't mistakenly capture byref in block nested in lambda

[Richard Smith]

Hi Ben,

On Tue, Jan 27, 2015 at 3:06 PM, Ben Langmuir <blangmuir at apple.com> wrote:

> Hi Richard,
>
> Could you take a look at this patch?  I haven’t messed with capturing in a
> long time and I don’t want to break it ;-)


LGTM as a diagnostics improvement (but see below).


>     Don't mistakenly capture byref in block nested in lambda
>
>     Even if an enclosing lambda captures a variable by reference, we
> capture
>     it by *copy* in a block unless the variable itself was declared with a
>     reference type. Codegen was already correct, but we were not diagnosing
>     attempts to modify the variable in Sema.
>

Hmm, I think I'm missing something here -- it looks like we would diagnose
it, but we'd use a suboptimal diagnostic: CheckForModifiableLvalue would
diagnose as err_typecheck_assign_const in this case.


>     This also fixes a crash on invalid when trying to modify a variable
> with
>     non-trivial copy constructor and assignment operator in such a
>     situation.
>

I can't reproduce this crash with current trunk, assuming this is supposed
to be covered by the uses of 'A' in test7 in SemaCXX/blocks.cpp.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20150128/f713ddf9/attachment.html>