Boost logo

Boost :

Subject: Re: [boost] [1.50.0] RELEASE BRANCH REOPENED (was: Beta schedule)
From: Beman Dawes (bdawes_at_[hidden])
Date: 2012-05-23 16:16:02


On Wed, May 23, 2012 at 9:45 AM, lcaminiti <lorcaminiti_at_[hidden]> wrote:
>
> Beman Dawes wrote
>>
>> On Wed, May 23, 2012 at 7:55 AM, lcaminiti &lt;lorcaminiti@&gt; wrote:
>>>
>>> Daniel James-3 wrote
>>>>
>>>> On 22 May 2012 23:46, Eric Niebler &lt;eric@&gt; wrote:
>>>>>
>>>>> Look for an email from us shortly. I'm taking it upon myself to reopen
>>>>> the release branch for BUG FIXES ONLY (no new libraries). Please get
>>>>> your changes in as soon as is convenient and prudent. It should be open
>>>>> for at least a week. Then we'll require release manager approval.
>>>>> Please
>>>>> be sure trunk tests are clean before merging anything, as usual.
>>>>
>>>> I added a 'release branch closed' event to the calendar for next
>>>> Monday, which is a little less than a week, but we normally do these
>>>> things on Mondays. This is in no way final (I just picked a date), and
>>>> might be changed later.
>>>>
>>>
>>> To confirm, does this mean that release is open for merging, including
>>> new
>>> features, until next Monday?
>>
>> "BUG FIXES ONLY" does not include new features.
>>
>> You can ask for permission to merge a new feature. If it has been
>> stable in trunk for awhile and is otherwise low risk, then we may OK a
>> merge. But we are playing catch up for a release that was supposed to
>> be done at the beginning of the month, so are trying to avoid anything
>> that will result in further delays.
>>
>
> Do I have permission to merge ScopeExit (improved), LocalFunction (new),
> Funcitonal/OverloadedFunction (new), and Utility/IdentityType (new)? If so,
> I can do that within today.
>
> I'm answering your questions below from another email to assess the risk.
>
>> It is an issue of risk. How long have these changes/refactorings been
>> stable in trunk?
> 1+ month.
>
>> How extensive were the changes?
> ScopeExit ("small" library) 20% new but all old regressions plus all new
> regressions pass.
> LocalFunction ("small/mid-size" library), OverloadedFunction ("small"
> library), and IdentityType ("tiny" library) 100% new.
>
>> Were the changes fragile or once they worked on your development platform,
>> did they pass all tests on other platforms?
> Regressions passed on all compilers with little efforts after they passed on
> MSVC and GCC on my development platform. Sun, and a little bit VACPP plus
> PGI, were the only compilers that required some amount of extra work. All of
> this was done 1+ month ago in trunk. That applies to all ScopeExit,
> LocalFunction, OverloadedFunciton, and IdentityType.
>
>> Have you done a local merge to release, and tested the results?
> Yes. I tested on MSVC 8.0, GCC 4.5.3 without and without C++11 feature
> (that's my development platform). The tests pass on my development platform
> for release as they do for trunk.
>
> Please advice.

That sounds OK to me; you have the risks pretty well covered. Several
other release managers are also in favor of a go ahead, so please
merge to release ASAP and keep a close eye on the release regression
tests.

Thanks,

--Beman


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk