NAME

docs/project/ticket_triaging.pod - Managing Tickets

DESCRIPTION

This document attempts to outline a set of best practices for dealing with tickets in Parrot's tracking system. It is targeted at Parrot developers and Ticket Wranglers and is not intended as advice or instruction for end users. Ticket filing procedures for end users are documented in docs/submissions.pod.

WHAT ABOUT TRAC?

Our preferred method of bug tracking at this point is trac: https://trac.parrot.org/

All Parrot developers are expected to pitch in and help keep the ticket tracker in a healthy state. This means you! Most of the document below still makes sense in terms of activities in trac, but the specifics are of course different with the new system.

Our previous bug tracking system was RT. In November 2009 all remaining RT tickets were closed, with many being reopened in Trac. No new issues should be opened in RT, but the old system is available at https://rt.perl.org. The Parrot issues are in the queue parrot.

TICKET HANDLING PROCEDURES

New Tickets

Where New refers to a pre-existing ticket sitting in the Parrot queue with a status of new.

Bug Triage

Involves deciding whether the ticket is a real bug, a feature request, a duplicate, or spam.

It is especially important to check that all new bugs which are marked [TODO], [PATCH], or [CAGE] really are bugs of the given class. This is because some bugs, such as [TODO] and [CAGE], get their status set to open to indicate that something should be done, rather than that someone is doing something.

Is this spam?
Assign the issue to the queue spam. Note that if this is successful, you will no longer have permissions to view the ticket.
Is this a duplication of an existing ticket?
Under the Action section, chose the "resolve as" option and select "duplicate" from the dropdown. Add a comment to the ticket along the lines of "Duplicate of Ticket #123".
Is there enough information?
If not, ask for more input.
Is it a [TODO] ticket?
Is the subject line in the format "[TODO] subsystem - issue?Change the status of [TODO] tickets to open to prevent them from appearing in a listing of new tickets.
Is it a [PATCH] ticket?
Is the subject line in the format "[PATCH] subsystem - issue?Make sure that there is actually a patch attached to the ticket. If you've applied a patch, be sure to include the revision in your response when closing the ticket.
* Is it a [CAGE] bug?
Is the subject line in the format "[CAGE] subsystem - issue?[CAGE] bugs should have their status changed to open to prevent them from appearing in a listing of new bugs.
* Assign the bug to someone if at all possible.

TODO Tickets

Claim ownership or interest ( CC ) of the ticket.
This way you will receive further correspondence about the ticket.
Run the test suite
make manitest
add the patch author to CREDITS or update the author's entry
add correspondence to the ticket stating that the patch was applied. Include the commit SHA1 in your response.
make sure that the ticket's 'Tag' includes 'Patch'
set the ticket's 'Patch status' to 'Applied'
set the ticket's 'resolve as' to 'resolved'

Old Tickets

If the ticket is more then 1 month old, then it's old.

Ping the requestor
Give the requestor at least 1 week to respond. If you receive no response, add a comment to the ticket saying that it is stalled because of no response from the requestor. Change the status to stalled.If it's a [PATCH] ticket, it's possible that the patch was applied but the ticket/patch status was never changed. Also, not all list traffic regarding a ticket ends up in the tracker. Look at the git repo to attempt to determine if the ticket was resolved.
Review of stalled tickets
Sometimes tickets are stalled because there's no hope if fixing them soon. Sometimes, no response is available from the requestor, or no one can verify the report. Review these tickets periodically. When possible, change their status to open or closed as appropriate.

Necessary Information

As alluded to earlier, tickets are much easier to resolve when they include sufficient information. For bugs, this is:

Specific error messages.
These can come from Parrot or the operating system. Copied and pasted messages are best; the exact wording is often important.
Platform details.
These include operating system and version, compiler information, processor architecture, and versions of included libraries. The file parrot_config.pasm may be useful.
Steps to reproduce.
At what point did you see the failure? Can you reproduce it? Do you have a small PIR program which demonstrates the problem?
Failure diagnostics.
Verbose diagnostics from prove -v -- including all error messages and diagnostics -- are often necessary to resolve test failures.
Backtraces.
Segfaults and other crashes within Parrot are much easier to resolve given a backtrace on a Parrot built with debugging symbols.

[PATCH] tickets for code almost always need tests and often need documentation. Feel free to ask the submitter to work on both, or to contact another committer for help in identifying and creating the appropriate tests and documentation.

Severity Guidelines

Occasionally the severity of a problem may govern how volunteers direct their resources to resolving tickets. Here are several criteria by which to determine the severity of a report.

Is there an exploitable security problem?
Can a malicious user destroy data, access sensitive information, or obtain undesired permissions to the rest of the system? If so, this is a high priority ticket.
Is there a crash from a pure-PIR program?
Users should never be able to crash Parrot writing normal PIR code. Such problems need high priority tickets.
Does the defect prevent a successful configuration, build, or installation of Parrot?
If Parrot cannot build, install, or run, the ticket has a high priority.
Does the defect cause test failures on a core platform?
All tests should pass on all core platforms in all releases of Parrot (as well as on trunk). Test failures, with the appropriate diagnostic information, are moderate priority.
Is the defect reproducable in the current development version?
Reproducable defects have a normal priority. Unreproducable tickets have a lower priority.
Does the defect affect a core platform?
Defects affecting non-core platforms have a lower priority (which reflects that we probably lack the expertise to deal with that platform).
Is there a test case suitable for the test suite?
A defect reported with a patch to the test suite (or a test case easily added to the test suite) may have a higher priority; it's much easier to diagnose and fix such problems.

TIPS FOR CORRESPONDENCE

Be Nice

Remember that every word you type into the ticket tracker is On The Record. Try not to say anything that could offend or hurt the feelings of anyone. That includes the ticket submitter and other developers. When, as a Parrot developer with commit rights, you send correspondence you are representing the Parrot project and, by proxy, The Parrot Foundation. If in doubt, either send the message privately or not at all.

Say thank you!

Try to add a little token of appreciation to every message you send in response to a ticket. Ticket requestors are doing labor for free! The least you can do is let them know that you appreciate their efforts.

Something like:

    Thanks,

    Thanks for following up.

    Thanks for reporting.

    Thanks for X!

... can work wonders. If you can make someone feel good about themselves maybe they'll submit another ticket/patch/whatever or perhaps some day become a Parrot developer.

Make it clear why the ticket status has changed

Always note why you're changing a ticket's status, particularly if you're closing or rejecting it. Nothing will irritate people more then letting them think that their ticket was unimportant or ignored.

Example Correspondence

    Hi,

    Can you retest for this ticket with the latest sources from git and confirm
    that this still an open issue?

    Thanks,

    -J

or

    Hi,

    Would you mind retesting with the latest sources from git?

    Thanks,

    -J

or

    Hi,

    Can you resubmit this patch against git master?

    Thanks,

    -J

or

    Patch applied as rXXX.  Thanks for submitting.

    -J

or

    No response for requestor.  Ticket being marked as 'rejected'.
    Thanks for reporting.

    -J

or

    This doesn't appear to be an issue anymore.
    Thanks for submitting.

    -J

or

    Marking this ticket as 'resolved' because it seems to have fixed itself.
    Thanks for following up.

    -J

TIPS FOR WRITING COMMIT MESSAGES

Put a subsystem identifier out the front
  [json]: commit message
If related to a Trac ticket, use the ticket number
  [json]: Resolve #731
Add a "Courtesy of <foo>" if supplied by someone else
  Courtesy of A. U. Thor <author@cpan.org>
Detailed commit messages are preferred
Make it clear what your intent is when committing. It makes future maintenance much easier.
  [PGE]:
  * Switched "PGE::Regex" to be "PGE::Grammar", to be more accurate.
  * Moved default rules from PGE::Regex into PGE::Match.
  * Updated various languages and tools to match.
Commit file names
You don't need to include the filename in the commit message as that's part of the commit itself. However, if your commit affects multiple directories, you may mention that, especially if it's part of a group of commits.
  [PDD07]: whitespace -- part 5
  ~ removed trailing spaces and tabs from t/exit/, t/dynpmc/, t/dynoplibs/
Group similar commits by parts
If all commits are much the same and require basically the same commit message, it can be useful to number the commit messages. For example:
  [tools]: smartlink functionality -- part 3
  ~ added regex attribute to Keyphrase class
  ~ filled in some more SmartLinkServer attribute init code
  ~ expanded LinkTree class functionality
  still TODO: merge smartlink and spec info, emit html, improve cmdline
  option code
You may optionally include items that are still todo, as it helps make your intentions clear.
More ideas
Look at past commit messages, and http://cia.navi.cx/stats/project/parrot for more best practices.