submitting patches that span projects, post repo merge

[Travis Crawford <tr...@gmail.com>]

Hey hadoop gurus -

With the merged repo, is the process for submitting patches that span
repos still one patch per repo? Is submitting a single patch with
changes to common, hdfs, and mapreduce acceptable post-merge?

If so, I can update the "Changes that span projects" section of
http://wiki.apache.org/hadoop/HowToContribute.


DETAILS

I finished a change for
https://issues.apache.org/jira/browse/HADOOP-7359, developed in a
branch on my github fork. Generating a diff between my branch and
trunk is straightforward, however, generating patches per-project
would require reworking the patch into 3 branches. Also, since the
patch changes something in common that's used in both HDFS & MR
submitting as separate patches will certainly break trunk for a brief
period.

With the merged repo, submitting patches that span projects via a
single diff is possible and potentially preferred. If
one-patch-per-project is still preferred I'll certainly do that, but
submitting a single patch is both easier from a development
perspective, won't cause temporary build breakage, and provides a
cleaner revision history.

Thoughts?

--travis

Re: submitting patches that span projects, post repo merge

[Todd Lipcon <to...@cloudera.com>]

The QA-bot unfortunately can still only deal with single-project patches. I
think there's an open JIRA to allow it to do cross-project tests.

I think for this patch, let's stick with the old way. We can do a single
commit when it's committed, but review-wise and QA-bot wise, separate JIRAs
is probably the way to go until we figure out a proper way to do these.

-Todd

On Sat, Jul 2, 2011 at 10:03 AM, Travis Crawford
<tr...@gmail.com>wrote:

> Hey hadoop gurus -
>
> With the merged repo, is the process for submitting patches that span
> repos still one patch per repo? Is submitting a single patch with
> changes to common, hdfs, and mapreduce acceptable post-merge?
>
> If so, I can update the "Changes that span projects" section of
> http://wiki.apache.org/hadoop/HowToContribute.
>
>
> DETAILS
>
> I finished a change for
> https://issues.apache.org/jira/browse/HADOOP-7359, developed in a
> branch on my github fork. Generating a diff between my branch and
> trunk is straightforward, however, generating patches per-project
> would require reworking the patch into 3 branches. Also, since the
> patch changes something in common that's used in both HDFS & MR
> submitting as separate patches will certainly break trunk for a brief
> period.
>
> With the merged repo, submitting patches that span projects via a
> single diff is possible and potentially preferred. If
> one-patch-per-project is still preferred I'll certainly do that, but
> submitting a single patch is both easier from a development
> perspective, won't cause temporary build breakage, and provides a
> cleaner revision history.
>
> Thoughts?
>
> --travis
>



-- 
Todd Lipcon
Software Engineer, Cloudera