[phobos] next release

[Brad Roberts]

On 2/14/2012 9:18 PM, Jonathan M Davis wrote:
> On Tuesday, February 14, 2012 21:07:06 Brad Roberts wrote:
>> On 2/14/2012 9:00 PM, Jonathan M Davis wrote:
>>> On Tuesday, February 14, 2012 20:54:43 Brad Roberts wrote:
>>>> Please do NOT start pulling changes into either druntime or phobos yet.
>>>> This last release was massive and there's a very real chance that we'll
>>>> need to do a fairly quick turn around release with regression fixes. 
>>>> Yes,
>>>> it could be done on a branch, but ...
>>>>
>>>> Even if there isn't a flurry of new regressions to fix, I believe it
>>>> worth a slide on other priorities to take the import and related
>>>> visibility changes through to completion and in relative isolation. 
>>>> Hopefully something like a 1-2 week release cycle instead of 2 months.
>>>>
>>>> Objections?
>>>
>>> I don't know. If we were going to be that quick about putting out another
>>> release just to fix the import issues, why didn't we just delay the 2.058
>>> release then?
>>>
>>> - Jonathan M Davis
>>
>> Not particularly relevant since it is released.  Do you have an objection to
>> the proposal that isn't based on altering the past? :)
> 
> I mean, if we were willing to put it out as-is rather than delay the release 
> long enough to deal with it, why not just do the next release like we normally 
> would and let it get fixed along with everything else rather than delay 
> everything while we fix it and then go through the whole release process again? 
> If the import stuff wasn't important enough to delay the release any further, I 
> don't see why it's important enough for us to rush another release just for 
> it.
> 
> - Jonathan M Davis

The last release had a ton of stuff in it blocked by the import changes and regressions from them.  It was worth taking
the parts that we knew (for a definition of knew that involves the confidence we gain through the test suites and a
short beta period) to be good and getting those fixes out the door rather than delaying another couple weeks trying to
finish the import stuff.

We have pull requests in-hand for most if not all of the import issues and I feel fairly confident that we can nail any
remaining issues reasonably quickly.  I expect that we (ok, read that as 'Walter') could get those pull requests in
tonight and/or tomorrow and spin a new beta nearly immediately.  Given a limited set of changes and a specific call for
testers to help regression test it better than most beta releases are tested, I expect that a complete release can be
turned around reasonably quickly.  We know that the import changes are one of the most request fixes/changes and that
it's  -- this change is arguably much more important and popular than most changes.

As to my point about regressions, I wasn't talking about the few that are in bugzilla and have been for a while, I'm
talking about the rather real probability that users picking up 2.058 will find new regressions.  I'm arguing that we
should hold off a brief period of time before destabilizing dmd with an onslaught of changes and another long release cycle.

If it feels better, you could pretend that tonight's release didn't actually happen and instead we held off a little
longer to debug the import changes, but I'm not a fan of self-delusion.

Anyway, it's just a suggestion based on experience from past releases and a really nice high demand feature with changes
in-hand.

Later,
Brad