In https://github.com/apple/swift/pull/7136 I tried to address some issues with
associated type inference's consideration of witnesses. A property of
associated type inference that sets it apart from other name lookup and
overload resolution problems is that the associated types needing inferenc
> On Jan 30, 2017, at 11:34 AM, Joe Groff wrote:
>
> In https://github.com/apple/swift/pull/7136 I tried to address some issues
> with associated type inference's consideration of witnesses. A property of
> associated type inference that sets it apart from other name lookup and
> overload res
Any recent changes to this code area, or new compiler bug?
Compile Swift Module 'PackageGraph' (4 sources)
Compile Swift Module 'Get' (9 sources)
Link Get
Valid decl has error type!
(subscript_decl "subscript(_:)" interface type='(Base.Index) ->
Base.Iterator.Element' access=public storage_kind=c
Old compiler bug.
Doug, it’s baack!
> On Jan 30, 2017, at 1:26 PM, Michael Ilseman via swift-dev
> wrote:
>
> Any recent changes to this code area, or new compiler bug?
>
> Compile Swift Module 'PackageGraph' (4 sources)
> Compile Swift Module 'Get' (9 sources)
> Link Get
> Valid
Over to Doug then
> On Jan 30, 2017, at 2:17 PM, Slava Pestov wrote:
>
> Old compiler bug.
>
> Doug, it’s baack!
>
>> On Jan 30, 2017, at 1:26 PM, Michael Ilseman via swift-dev
>> mailto:swift-dev@swift.org>> wrote:
>>
>> Any recent changes to this code area, or new compiler bug?
I saw one of these before Joe’s change went in. It might be more
non-determinism with archetypes...
> On Jan 30, 2017, at 3:11 PM, Michael Ilseman via swift-dev
> wrote:
>
> Joe, could this be due to your changes?
>
> Assertion failed: ((inContext->isEqual(repInContext) || inContext->hasError
Joe, could this be due to your changes?
Assertion failed: ((inContext->isEqual(repInContext) || inContext->hasError()
|| repInContext->hasError()) && "Potential archetype mapping differs from
representative!"), function operator(), file
/Users/buildnode/jenkins/workspace/oss-swift-incremental-R
> On Jan 30, 2017, at 3:19 PM, Slava Pestov wrote:
>
> I saw one of these before Joe’s change went in. It might be more
> non-determinism with archetypes…
I’m working on the general problem here.
- Doug
>
>> On Jan 30, 2017, at 3:11 PM, Michael Ilseman via swift-dev
>> mailto:swift
