On 9/23/14, 9:27 AM, David Nadlinger wrote:
> On Tuesday, 23 September 2014 at 16:20:25 UTC, Andrei Alexandrescu wrote:
>> I guess you can disable the whole feature on Win64 and leave it to
>> someone else to introduce it. You can't work on this stuff without a
>> box. -- Andrei
>
> Yes, but don't forget that there are still a few actual, unresolved
> issues with the PR (and I'm not talking about my documentation nit-picks
> here, they could also be done separately).

Sean? -- Andrei

On Tuesday, 23 September 2014 at 16:38:14 UTC, Andrei Alexandrescu wrote:
> On 9/23/14, 9:27 AM, David Nadlinger wrote:
>> On Tuesday, 23 September 2014 at 16:20:25 UTC, Andrei Alexandrescu wrote:
>>> I guess you can disable the whole feature on Win64 and leave it to
>>> someone else to introduce it. You can't work on this stuff without a
>>> box. -- Andrei
>>
>> Yes, but don't forget that there are still a few actual, unresolved
>> issues with the PR (and I'm not talking about my documentation nit-picks
>> here, they could also be done separately).
>
> Sean? -- Andrei

I wasn't aware of any.  Someone suggested moving Generator to a separate module and I explained why this isn't advisable.  If there are other issues, I would appreciate it if someone would restate them.

On Tuesday, 23 September 2014 at 16:56:36 UTC, Sean Kelly wrote:
> I wasn't aware of any.  Someone suggested moving Generator to a separate module and I explained why this isn't advisable.  If there are other issues, I would appreciate it if someone would restate them.

1) There is still Martin's point that Boost.Coroutine seems to use a different design. I haven't had a look at it in any detail, but I think it would be good to briefly comment on why their choices would not be a good fit for us. After all, one would suspect that they started out with vaguely similar design space restrictions and goals.

2) There are still a couple of typos in the docs, and at least one code quality suggestion (`auto ref`). Just scroll through the non-outdated GitHub comments.

David

On Tuesday, 23 September 2014 at 17:31:31 UTC, David Nadlinger wrote:
> On Tuesday, 23 September 2014 at 16:56:36 UTC, Sean Kelly wrote:
>> I wasn't aware of any.  Someone suggested moving Generator to a separate module and I explained why this isn't advisable.  If there are other issues, I would appreciate it if someone would restate them.
>
> 1) There is still Martin's point that Boost.Coroutine seems to use a different design. I haven't had a look at it in any detail, but I think it would be good to briefly comment on why their choices would not be a good fit for us. After all, one would suspect that they started out with vaguely similar design space restrictions and goals.

I thought I did :-)  Generator is a limited version of a coroutine, which I created both in response to some features in other languages like Go and because of something that came up during Chuck Allison's talk at the conference this year.  It's possible that Boost-style coroutines will be added later and Generator will be made a special case of these, but that should not affect the current API.  I think it still has merit as a standalone thing, as Boost-style coroutines won't work as an input range (which is the point of Generators).


> 2) There are still a couple of typos in the docs, and at least one code quality suggestion (`auto ref`). Just scroll through the non-outdated GitHub comments.

I must have missed this.  I'll check.

On Tuesday, 23 September 2014 at 08:06:36 UTC, Idan Arye wrote:
> On Tuesday, 23 September 2014 at 03:07:41 UTC, Jet wrote:
>> I hope the Dlang can have the Micro-thread at the language-level. Like the Goroutine, maby.
>
> Sure it can - it's called "Fibers":
>
> http://dlang.org/phobos/core_thread.html#.Fiber

Oh, language-level ?

On Wednesday, 24 September 2014 at 11:55:46 UTC, Jet wrote:
> On Tuesday, 23 September 2014 at 08:06:36 UTC, Idan Arye wrote:
>> On Tuesday, 23 September 2014 at 03:07:41 UTC, Jet wrote:
>>> I hope the Dlang can have the Micro-thread at the language-level. Like the Goroutine, maby.
>>
>> Sure it can - it's called "Fibers":
>>
>> http://dlang.org/phobos/core_thread.html#.Fiber
>
> Oh, language-level ?

What's wrong with a library implementation?

On 9/23/14 12:11 PM, Sean Kelly wrote:
> On Tuesday, 23 September 2014 at 15:44:19 UTC, Andrei Alexandrescu wrote:
>> On 9/23/14, 6:41 AM, Sean Kelly wrote:
>>> On Tuesday, 23 September 2014 at 12:19:52 UTC, Daniel Kozak via
>>> Digitalmars-d wrote:
>>>>
>>>> I know, but I mean there is no scheduler in standard library or at
>>>> language-level
>>>
>>> That code has been written for almost a year now.  Someone will pull it
>>> eventually :-/
>>
>> In fairness,
>> https://github.com/D-Programming-Language/phobos/pull/1910 needs your
>> attention. It's failing unittests. -- Andrei
>
> On Win64 only.  The problem isn't with the pull request but rather with
> fibers on Win64 or with something else entirely (unclear, given the
> stack traces I've seen so far).  ie. The problem was that I added
> unittests that didn't exist before which may have uncovered an existing
> issue.  I've considered simply disabling the tests on Win64 until
> whatever is going on there can be sorted out.  I don't have a Win64 test
> environment so I've been relying on others to provide info.

I have a win 64 VM running on VMWare, which I use for work, but not super-often.

Send me an email, I'll give you access via TeamViewer if you want to try stuff out.

-Steve