asterixdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Till Westmann" <ti...@apache.org>
Subject Re: Regarding git workflows and who must push a patch
Date Tue, 05 May 2015 07:04:19 GMT
This is indeed a very interesting question.
It's not clear to me, how the push records are being recorded and if we 
could achieve the same QoS e.g. by having the same access control on the 
Gerrit pushes.

Also, the git page [1] says that "Setting up gitpubsub should provide 
sufficient flexiblity ...", so I'm wondering if there's a way to use 
that to provide pre-commit testing.

Do our mentors have so experience with this?

Thanks,
Till


[1] https://www.apache.org/dev/writable-git

On 4 May 2015, at 12:19, Ian Maxon wrote:

> Hi,
> Right now our git workflow uses Gerrit as a means to facilitate code 
> review
> and verification of patches. The way we have it set up is that the 
> code
> must be reviewed, and must have 'mvn verify' pass on our Jenkins 
> server (or
> have that overridden by manual testing, but this shouldn't happen
> normally). Patches get submitted to Gerrit first, and then when merged 
> into
> Gerrit's internal master after review, they get pushed to "official"
> repositories that folks fetch changes from normally.
>
> I was reading about how to alter this workflow a bit to instead now 
> use
> ASF's git repository (instead of Google code) as a new upstream from 
> our
> Gerrit server, and I came across this document:
> https://www.apache.org/dev/writable-git . The concerning part to me, 
> was
> this: "The ASF repo must be the canonical master repo that all 
> committers
> push changes to." In the explanation of this rule, this page(
> http://www.sunstarsys.com/essays/git-and-non-repudiation) is linked, 
> and it
> goes on to explain the concept of push records. If I understand the 
> two
> above pages correctly, the situation is that only a committer (i.e. , 
> not a
> robot authorized by a committer ) can actually log into the ASF git 
> repo to
> push new changes. Is this indeed correct? If so what options might we 
> have,
> for a git workflow similar to the one we have now? The reason we 
> deployed
> Gerrit and Jenkins was to have better gatekeeping over what commits 
> got
> into master, so I think any tool chain that would achieve that end 
> could be
> worth discussing. Do Github pull requests work on ASF mirrored repos, 
> the
> same as they do on normal Github repos?
>
> Thanks,
> - Ian

Mime
View raw message