[PATCH] [Scalarizer] Fix potential for stale data in Scattered across invocations

[Stephen Hines]

Ping. This is a pretty simple patch for a particularly nasty bug in the
scalarizer.

Thanks,
Steve

On Fri, Jun 19, 2015 at 4:28 PM, Stephen Hines <srhines at google.com> wrote:

> Ping. Matt added a test that fails with the implicit preconditions of the
> pass (i.e. that gathered and scattered both start as empty - this is now an
> assertion in the code as well). It also fixes the issue with returning
> false when changes have indeed been made to the underlying code.
>
> Thanks,
> Steve
>
> On Mon, Jun 15, 2015 at 1:53 PM, Matt Wala <wala at google.com> wrote:
>
>> Hi Chandler:
>>
>> I apologize, I did not mean to break any rules.
>>
>> On Fri, Jun 12, 2015 at 5:50 PM, Chandler Carruth <chandlerc at google.com>
>> wrote:
>>
>>> Matt, there are a bunch of problems here.
>>>
>>> First, Steve Hines accepted this revision in Phabricator, but didn't
>>> actually send any LGTM or other comment to the mailing list, so it didn't
>>> look like *any* review happened.
>>>
>>> Second, Steve, I'm not sure how familiar with this code you are as I've
>>> never seen you really contribute to it, so it might be better to let others
>>> review the patch.
>>>
>>> Third, making the point that sufficient review had not happened, this is
>>> a bugfix patch which does not contain a test case. I feel confident that a
>>> dangling pointer style bug could be exercised with a good test case and
>>> ASan. We have ASan bots, so having the test case in the tree is really
>>> useful.
>>>
>>> Matt, please update the review thread when you have added such a test
>>> case.
>>>
>>
>> I've opened a new review with a test case attached.
>>
>> Here's the link to the new review: http://reviews.llvm.org/D10459
>> <https://urldefense.proofpoint.com/v2/url?u=http-3A__reviews.llvm.org_D10459&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=Tv1U250hSfPBgwWqFA2MaisotfFOWTMJeZ7j6rHGHVU&s=A50ZEB2MlBI7--kMpMuxyF8aLVMuNmUEjFq9dk6m-fA&e=>
>>
>> Matt
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150629/e6731723/attachment.html>