Hmm... Good questions. I assumed we'd do a 5.1.2 pretty quickly.
The time between the minor branches is waaayyyy too long.

I think might be fairly simple: Bug fixes have to be able to land somewhere. 
And that somewhere should not be tied to new features.

So, until we have a release on the next minor version we do patches on the 
current minor release.
I.e. we have no 5.2 release - and frankly who knows when we'll have that. So 
until then we should do 5.1.x releases.

Once we have a 5.2.x release, it's up to whether we'll find a maintainer for 
the 5.1 branch.

Not sure I can trust jira now, but we currently have 16 5.1.2 changes. Might be 
time to do a patch release again.
(Some are somewhat important for the Trino integration)

Here's a query that finds issues against 5.2.0 that are also in 4.16.1, 4.16.2, 
or 4.17.0 but not in 5.1.2 or 5.1.1.

https://issues.apache.org/jira/issues/?jql=project%20%3D%20PHOENIX%20AND%20fixVersion%20%3D%205.2.0%20and%20(fixVersion%20%3D%204.16.1%20or%20fixVersion%20%3D%204.16.2%20or%20fixVersion%20%3D%204.17.0)%20and%20fixVersion%20!%3D%205.1.2%20and%20fixVersion%20!%3D%205.1.1

Returns 17 issues for me, including 10 marked as resolved.

PHOENIX-6457    Geoffrey Jacoby    4.17.0, 5.2.0    Reopened    Actions
PHOENIX-6454    Swaroopa Kadam    4.17.0, 5.2.0    Open    Actions
PHOENIX-6451    Richárd Antal    4.17.0, 5.2.0    Resolved    Actions
PHOENIX-6447    Sandeep Pal    4.16.1, 4.17.0, 5.2.0    Resolved    Actions
PHOENIX-6444    Rushabh Shah    4.17.0, 5.2.0    Resolved    Actions
PHOENIX-6437    Ankit Jain    4.17.0, 5.2.0, 4.16.2    Resolved    Actions
PHOENIX-6435    Xinyi Yan    4.16.1, 4.17.0, 5.2.0    Resolved    Actions
PHOENIX-6430    Jacob Isaac    4.17.0, 5.2.0    Resolved    Actions
PHOENIX-6429    Jacob Isaac    4.17.0, 5.2.0    Resolved    Actions
PHOENIX-6420    Tanuj Khurana    4.17.0, 5.2.0    Resolved    Actions
PHOENIX-6399    Istvan Toth    4.17.0, 5.2.0    Resolved    Actions
PHOENIX-6397    vikas meka    4.17.0, 5.2.0    In Progress    Actions
PHOENIX-6395    vikas meka    4.17.0, 5.2.0    In Progress    Actions
PHOENIX-6379    vikas meka    4.17.0, 5.2.0    Resolved    Actions
PHOENIX-6227    Geoffrey Jacoby    4.17.0, 5.2.0    Open    Actions
PHOENIX-6085    Richárd Antal    4.17.0, 5.2.0    Patch Available    Actions
PHOENIX-5346    Unassigned    4.17.0, 5.2.0    Reopened    Actions

Maybe the owners can have a quick look.

Cheers.

-- Lars

On Wednesday, May 12, 2021, 10:42:58 PM PDT, Istvan Toth 
<[email protected]> wrote: 





I've also wanted to bring this up after 5.1, but forgot while dealing with
the problems in 5.1.0

Do we want to maintain patch branches ?
For how long ?
What do we want to commit to the maintenance branches ?
Who commits to the maintenance branches ?

To kickstart the discussion, here are my current assumptions, which are of
course colored by the requirements at $dayjob:

Do we want to maintain patch branches ?

I think we should. Publishing patch releases that focus on bug fixes
(and supporting new HBase releases) and help users in a big way.
I believe that both large contributors maintain internal branches.
Doing most of that work on the apache maintenance branches shouldn't be a
lot of additional net effort.

How long should we maintain patch branches ?

I feel that maintaining them until we start the next minor release process
is
a good compromise.
Of course this also assumes that we also make semi-regular patch
releases on a 1-3 month cadence.
Getting into the habit of doing a "final" patch release around the
release of the next minor version would also be helpful for users.

What do we want to commit to the maintenance branches ?

NOTHING that breaks compatibility
EVERY significant bug fix (for some definition of significant)

For everything else, I personally make the decision based on effort:
If it's a change that applies cleanly, and requires no additional work, I
generally backport it.

Who commits to the maintenance branch ?

I like the current practice that the original committer should generally do
it.
Before starting the patch release RC process, the RM can compare the main
branch and the maintenance branch, and backport the
changes that the RM deems necessary.

I have also been operating on the assumption that while the RC is open,
only the RM should commit to the maintenance branch,
but that's just me not wanting to interfere with Xinyi's work, and I'd be
happy either way.

regards
István


On Wed, May 12, 2021 at 10:21 PM Geoffrey Jacoby <[email protected]> wrote:

> Have we actually decided that branch-5.1 will produce our next version of
> Phoenix?
>
> I've been assuming that 4.16 (and 5.1, which I'll admit I forgot existed)
> were just for bug fixes, and that the next significant version of Phoenix
> would be 5.2 / 4.17 sometime in H2 of 2021. (Or potentially just 5.2, if
> the "retire 4.x" proposal on another thread passes.)
>
> We already have several changes in 4.x and master that add columns to
> System.Catalog and hence can't go out in 4.16.2 or 5.1.2. (See PHOENIX-6247
> -- which should be marked Resolved -- and PHOENIX-6457).
>
> Do we need the dual maintenance of a 5.1 patch branch?
>
> Geoffrey
>
> On Wed, May 12, 2021 at 12:55 PM [email protected] <[email protected]>
> wrote:
>
> > Hi all,
> >
> > I noticed some changes that went into both master and the 4.x branch, but
> > not into branch-5.1.
> >
> > Unfortunately I do not have time to check each edit.
> >
> > When you commit changes, please do not forget about branch-5.1, which
> will
> > produce our next version of Phoenix.
> >
> > Cheers.
> >
> > -- Lars

> >
>


-- 
*István Tóth* | Staff Software Engineer
[email protected] <https://www.cloudera.com>
[image: Cloudera] <https://www.cloudera.com/>
[image: Cloudera on Twitter] <https://twitter.com/cloudera> [image:
Cloudera on Facebook] <https://www.facebook.com/cloudera> [image: Cloudera
on LinkedIn] <https://www.linkedin.com/company/cloudera>
<https://www.cloudera.com/
>
------------------------------

Reply via email to