[cfe-dev] Problem with DW_AT_abstract_origin references

[David Blaikie]

Great - thanks for letting us know/reporting this!

(even though it's already fixed - I'm gratified to know it was a real
user-impacting bug that I found/fixed by inspection ;))

On Fri, Jul 25, 2014 at 7:38 AM, Vladimir Nikulichev <nvs at tbricks.com> wrote:
> On Jul 24, 2014, at 7:43 PM, David Blaikie <dblaikie at gmail.com> wrote:
>
>> Yep - that'd be a known bug in Clang 3.4 (well, known when I looked at
>> the code, saw that it was no good, and fixed it) that has, hopefully,
>> since been fixed. At least I think that's what you're seeing.
>>
>> Have you tried reproducing this with Clang ToT? (or a 3.5 release candidate?)
>>
>> You should be able to reproduce this by writing a small example with
>> __attribute__((force_inline)) (or is it always_inline, I forget...
>> maybe it's one word, without the underscore... ), I would imagine -
>> even without LTO.
>>
>> Anyway, if you have some preprocessed source, I can compile it with
>> Clang ToT and show you the dwarfdump to see if it makes more sense
>> (but I can't easily verify that GDB doesn't infinitely recurse without
>> a fully linkable/executable example).
>>
>> - David
>
> Yes, it works fine in 3.5.