Alan D. Cabrera wrote:
John Sisson wrote:
One of the issues I see with the current process we have for changes
under RTC is that it is hard to keep track of what patches are
pending RTC.
Ken suggested that we reintroduce the STATUS file as a way of keeping
track of the status of patches (
http://www.mail-archive.com/dev%40geronimo.apache.org/msg24780.html ).
On the same thread, Dain suggested introducing a "review-required"
status in JIRA (
http://www.mail-archive.com/dev%40geronimo.apache.org/msg25122.html )
and is the method of tracking work that I prefer.
PROPOSAL
1. Add a "review-required" and "review-complete" statuses to JIRA. I
thought having two statuses might allow a cleaner workflow in JIRA,
but would be interested in hearing others opinions.
I think that we only need a Review status. But, in any case, Jira is
definitely the way to go.
2. To make it easy for those reviewing and voting on the patches
(there could end up being a number of revisions of the patch before
it is accepted) the file names of the patches attached to the JIRA
should be prefixed with the JIRA issue identifier followed by an
optional text followed by a mandatory patch version number (starting
at 1).
Example patch names:
GERONIMO-1234-FixNPE-v1
GERONIMO-1234-FixNPE-v2 (second attempt at patch)
GERONIMO-3421-v1
Why should a patch that has been attached to a specific Jira issue
include the number of the jira issue? I don't mind but it seems odd
and usually that means that I am misunderstanding something.
I was just trying to make it simpler for people so there is no confusion
with what patch file they are working with once they save the patch to
their PC from the JIRA issue.
2.1 This status should only be set by a committer (can we can get
JIRA to enforce this?) when they have tested the patch attached to
the JIRA and believe it is ready for review.
There should be a way to enforce it.
2.2 The JIRA should contain all information about the patch. If the
changes were previously discussed on the dev list prior to the JIRA
being created, a summary of the discussions should be moved into the
JIRA so that those reviewing the patch have all the information in
one place. It would also be preferable to add links to the original
discussions on the dev list archives. The way we document changes
may be subject to change (e.g. detailed documentation placed in a
linked JIRA) based upon the outcome of the discussions in the thread
"[DISCUSS] Tracking documentation tasks in JIRA ( was Re: [RTC]
Clarification please from the PMC )"
2.3 Each PMC member who reviews the patch attached to the JIRA must
do the following:
* Add a JIRA comment containing the file name of the patch they
reviewed. This is so there is no confusion if there ends up being
multiple revisions of the patch when collating votes.
* In the JIRA comment add the results of their review (e.g.
comments or a vote). If a PMC member vetos the patch, they must
include a technical justification in their JIRA comments. I propose
that when there is a veto that we leave the status as
"review-required", as others may still want to vote and so that the
patch remains getting daily visibility in the "JIRAs Pending Review"
daily email (proposed below). The committer can then re-submit
another patch (where the patch filename has the version number bumped
up)
* If a PMC member is the person who completes the vote ( three
binding +1s and no vetos) for the latest version of the patch then
they should change the status to "review-complete".
Just need to move it on to the regular Open status, no?
Thought it may be clearer to have a different status showing the review
is complete and have the JIRA workflow match the true workflow. For
example if we were to move to open status after the review is complete,
then the user would be given the opportunity to go back to
review-required status, which isn't really a valid state transition, but
maybe we should keep it simple and rely upon common sense? Any JIRA
experts out there who can comment on the benefits/disadvantages of
having an additional review-complete status?
John
3. Non-committers who submit patches will not be able to set this
status. A committer needs to pick up their patch and test it
(possibly making changes to the patch). When the patch is ready the
committer then sets the "review-required" status.
4. Have a daily email automatically sent to the dev list containing
JIRA's pending review. It appears this should be easy to implement
as it would be a variation of the weekly "Unassigned Patches" reports
that are currently in place.
I would be interested in your comments Jason, as you are more
familiar with customizing JIRA.
If this proposal is accepted I will document it as part of the work I
plan to do to document the use of JIRA in
http://issues.apache.org/jira/browse/GERONIMO-2080 .
This stuff sounds pretty cool.
Regards,
Alan