[PATCH] Linker: Replace overridden subprograms

[Duncan P. N. Exon Smith]

r224389. 

Sorry, I'd interpreted your previous comment as an LGTM :).

> On 2014-Dec-16, at 15:41, Rafael Espíndola <rafael.espindola at gmail.com> wrote:
> 
> LGTM
> 
> On 16 December 2014 at 17:55, Duncan P. N. Exon Smith
> <dexonsmith at apple.com> wrote:
>> 
>>> On 2014 Dec 16, at 14:38, Rafael Espíndola <rafael.espindola at gmail.com> wrote:
>>> 
>>>> It's asymmetrical (asymmetry happens all the time in linkers, right?), but
>>>> I'm not sure there's a good way to fix this without invasive changes to
>>>> `MapValue()`.  I'd rather leave "solving this correctly" until we have a
>>>> custom class for subprograms (as mentioned in PR21910).
>>> 
>>> So the logic is asymmetrical, but the end result is normally
>>> symmetrical, which is why some tests use
>>> ; RUN: llvm-link %p/foo.ll %p/Inputs/foo.ll ...
>>> ; RUN: llvm-link %p/Inputs/foo.ll %p/foo.ll ...
>>> 
>>> But yes, this can be done in another patch. Please just leave a FIXME
>>> in the code somewhere.
>> 
>> Done.
>> 
>>>> (I wouldn't even be trying to solve this yet if it weren't blocking
>>>> our Release+Asserts bootstrap.)
>>> 
>>> I think the patch is OK with the SrcM->getNamedMetadata("llvm.dbg.cu")
>>> check removed.
>> 
>> Attached.
>> 
>> @David, does this look okay to you too?  (I won't be closing PR21910; I
>> figure we can use that to track a better long-term fix.)
>>