[PATCH] Fix unconditional dereference of a WeakVH in CGDebugInfo TypeCache

[David Blaikie]

On Tue, Mar 4, 2014 at 1:39 PM, Reid Kleckner <rnk at google.com> wrote:

> Added a test case for this in r202891.  It boiled down to using
> -gline-tables-only.  You might want to change the code to do less work if
> we're not actually going to emit type information, though.
>

Oh, thanks a bunch! That's good to know... yeah, we should probably have a
lot more early-outs so we don't end up doing this sort of work in line
table only mode.


>
>
> On Tue, Mar 4, 2014 at 10:42 AM, David Blaikie <dblaikie at gmail.com> wrote:
>
>>
>>   LGTM - though I wonder whether we should actually be using TrackingVH
>> here (not sure when we're expecting to "lose" types like this).
>>
>>   A specifically reproducible test case might be helpful to file as a "to
>> investigate" sort of thing. (where we can replace the 'if' you're adding
>> with an assert, then reduce the test case, etc and better understand what's
>> going on here and why we end up in this situation)
>>
>> http://llvm-reviews.chandlerc.com/D2944
>>
>> BRANCH
>>   fix-weak-dbg-info
>>
>> ARCANIST PROJECT
>>   clang
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140304/19761000/attachment.html>