[PATCH] D24595: [lib/LTO] Remove now unneded hack for undefined symbols

Teresa Johnson via llvm-commits llvm-commits at lists.llvm.org
Thu Sep 15 10:17:42 PDT 2016


On Thu, Sep 15, 2016 at 9:44 AM, Mehdi Amini <mehdi.amini at apple.com> wrote:

>
> On Sep 15, 2016, at 9:35 AM, Rafael Espíndola <rafael.espindola at gmail.com>
> wrote:
>
> BTW, that doesn't seem to be the only issue.
>
> Right now patatino gets internalized, which fails the verifier, that would
> be fixed by  pr30396 or some workaround.
>
> But, adding -upatatino to the command line causes gold to fail because foo
> is internalized. I can't quite make what is the code path that is missing,
> but nothing creates the llvm.compiler.used that should prevent foo from
> being removed.
>
>
> Indeed, it looks like the call to updateCompilerUsed got lost in the new
> API…
>
> Teresa: for ThinLTO it got lost when you moved to the index-based
> internalization in http://reviews.llvm.org/D20290 I believe.
>

Do we need that for ThinLTO? Note the AsmUndefinedRefs are being preserved
via the MustPreserveGV callback thinLTOInternalizeModule.

Teresa


>
>> Mehdi
>
>
>
>
> Cheers,
> Rafael
>
>
>
>
> On 15 September 2016 at 11:27, Rafael Espíndola <
> rafael.espindola at gmail.com> wrote:
>
>>
>> Yes that sounds the proper fix: we should have llvm-nm report the correct
>>> list for instance.
>>>
>>
>> Reported https://llvm.org/bugs/show_bug.cgi?id=30396 to track it.
>>
>> Cheers,
>> Rafael
>>
>>
>
>
>


-- 
Teresa Johnson |  Software Engineer |  tejohnson at google.com |  408-460-2413
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160915/51233fad/attachment.html>


More information about the llvm-commits mailing list