[llvm] r257915 - Don't try to check all uses if lazy loading.

[Mehdi Amini via llvm-commits]

> On Jan 19, 2016, at 11:40 AM, Hans Wennborg via llvm-commits <llvm-commits at lists.llvm.org> wrote:
> 
> On Tue, Jan 19, 2016 at 11:14 AM, Rafael Espíndola
> <rafael.espindola at gmail.com> wrote:
>> On 19 January 2016 at 14:06, Hans Wennborg <hans at chromium.org> wrote:
>>> Should this and r257920 be merged to 3.8?
>> 
>> Your call.
>> 
>> This patch might fix a bug in some other linker using the lto api and
>> lazy loading.
>> 
>> r257920 is just an extra assert, so we can just let it bake a bit
>> longer and wait for 3.9.
> 
> OK, doesn't sound like it was super important. I'll skip merging.

So unfortunately this was quite important, for a not-trivial reason:

The asserts were introduced in: r256105, *before the 3.8 branch*.
The asserts were reverted in r257751, http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20160111/324977.html
because : https://llvm.org/bugs/show_bug.cgi?id=25919
The revert was apparently after the branching, but this wasn't cherry-picked into the release.

So this commit is a bugfix, for a bug that prevents from bootstrapping LTO in 3.8. It seems we are lacking bots to validate the release, this is seems quite unfortunate...

Do we have a 3.8.1 planned? This should definitely be cherry-picked.

-- 
Mehdi