I’d be OK with that.

I think we should firm up a plan at 
https://issues.apache.org/jira/browse/CALCITE-1717 
<https://issues.apache.org/jira/browse/CALCITE-1717> before we act. Let’s 
continue discussion there.

Do you think we need a formal vote for this change? Repositories are pretty 
fundamental to the integrity of the project, so I’m inclined to think we do.

Julian



> On Mar 21, 2017, at 11:37 AM, Josh Elser <[email protected]> wrote:
> 
> Michael's latter suggestion is what I was thinking about :)
> 
> I've gone ahead and put in the request for an Avatica repo. We can do
> whatever we want with it then.
> 
> On Mon, Mar 20, 2017 at 6:11 PM,  <[email protected]> wrote:
>> You can use git subtree split to extract the Avatica commits into a new 
>> repository. This would maintain the history but would generate new commit 
>> IDs. Alternatively the history could start with the entire Calcite history 
>> with the first commit being one which deletes all non-Avatica files.
>> 
>> --
>> Michael Mior
>> [email protected]
>> 
>> From: Julian Hyde
>> Sent: Monday, March 20, 2017 18:09
>> To: [email protected]
>> Subject: Re: Commits during release vote
>> 
>> I’m not sure what are the options for splitting a git repo. Do you think 
>> that the avatica repo will retain the commit ids of the current repo? 
>> (Presumably to do that we need to keep all of the calcite commits, then 
>> delete the calcite files.) Or do you think the new repo should start from an 
>> empty history and contain only avatica commits?
>> 
>> I’ve logged https://issues.apache.org/jira/browse/CALCITE-1717 
>> <https://issues.apache.org/jira/browse/CALCITE-1717> to track.
>> 
>> We could also consider, at some point, asking for an AVATICA prefix for JIRA 
>> cases.
>> 
>> Julian
>> 
>>> On Mar 20, 2017, at 2:12 PM, Josh Elser <[email protected]> wrote:
>>> 
>>> How fitting :)
>>> 
>>> It should be fairly simple -- let me put in a request to infra. Once we 
>>> have a repo, it should be self-service for us to move the stuff over there.
>>> 
>>> Julian Hyde wrote:
>>>> Thanks - I’ll rebase after the release.
>>>> 
>>>> I’m sitting in a cafe right now and there’s a guy right outside the window 
>>>> re-potting the plant pots: pull up each plant individually, pull its roots 
>>>> apart from other plants, and put into a new pot with more room around it 
>>>> to grow.
>>>> 
>>>> We should definitely do that with Calcite/Avatica. If someone has the time 
>>>> and energy to move Avatica into a separate repo I will support that.
>>>> 
>>>> Julian
>>>> 
>>>> 
>>>>> On Mar 20, 2017, at 9:34 AM, Josh Elser<[email protected]>  wrote:
>>>>> 
>>>>> Oops, sorry for causing some pain, Julian. I'm ok with a 
>>>>> rebase/force-push. It completely slipped my mind that I should have held 
>>>>> off on those commits :)
>>>>> 
>>>>> I still think separating out Avatica from the Calcite repo would be 
>>>>> better long term (Shi had gotten confused over this, recently), but, like 
>>>>> you said, it's a pretty minor thing.
>>>>> 
>>>>> Julian Hyde wrote:
>>>>>> Josh,
>>>>>> 
>>>>>> I see you made 3 commits to Calcite’s master branch yesterday to fix 
>>>>>> Avatica issues. There is a release vote for Calcite underway, and the 
>>>>>> last few changes for that are in branch-1.12 but not in master. Up to 
>>>>>> this point we have been able to keep the commit tree linear (i.e. there 
>>>>>> are no merge commits with two parents), and I would like to retain that.
>>>>>> 
>>>>>> To keep the tree linear, I will need to rebase master onto branch-1.12 
>>>>>> after the release, then do a force push to master. Is that OK with you?
>>>>>> 
>>>>>> Devs,
>>>>>> 
>>>>>> Going forward, there are several solutions (lock the repository during 
>>>>>> release votes, force push to master after a a release, and allow the 
>>>>>> repository to have merge commits). And splitting the Calcite and Avatica 
>>>>>> repositories would at least allow Avatica development to continue during 
>>>>>> Calcite release votes. We should discuss the various options. But in my 
>>>>>> opinion, we are still small enough that the current policy (pausing 
>>>>>> commits during release votes, to keep a linear release history) is 
>>>>>> working fine.
>>>>>> 
>>>>>> Julian
>>>>>> 
>>>> 
>> 
>> 

Reply via email to