Triage for Bugzilla

Expectations

All teams working on Firefox using either or both Mozilla-central and Bugzilla are expected to follow the following process.

Components

You should understand the list of components that your team is responsible for. Each component must have a Triage Owner. (File a bug to update a component’s Triage Owner, or see this sheet in order to set up a triage rotation).

Triage Owners

Triage Owners are responsible for ensuring that bugs are triaged within the expected timeframe, as well as acting as the primary point of contact for triage related questions. While it’s their responsibility to ensure triage happens, it doesn’t necessarily mean only they can or should perform triage.

A good starting point is for anyone senior enough in the team to triage bugs as they are filed, leaving bugs that need discussion untriaged. Then schedule a weekly triage meeting to discuss and triage bugs where required.

Triage

Incoming bugs could be serious and we may need to react quickly. Users that have invested the time to inform us of a bug would like to feel that we are listening.

  • All new bugs should be quickly assessed on a daily basis to ensure that security bugs can be actioned quickly.

  • All new bugs should be fully triaged, or under active investigation, within one week of being created.

Important bugs monitoring and fixing

Bug severities are there for a reason. Tracked bugs must be taken particularly seriously, too. Regressions are important to users, regressed functionality is an indication that the product is getting worse not better. They should, best-case, never hit release but be fixed in beta.

  • All new S1 and sec-critical bugs should get the full attention of anyone that can reasonably help

  • All new S2 and sec-high bugs should be assigned (with caveats) and monitored closely (weekly team meeting)

  • Tracked bugs should be monitored closely (weekly team meeting)

  • Regressions should be monitored closely (weekly team meeting)

All of this can be found on the “Important” tab of https://bugdash.moz.tools/ (after selecting your components).

*::General components

We can’t expect users to know the details of our component structure so we may need to help routing bugs to the right place.

  • Some teams have *::General components. New bugs in these components should be triaged into more specific components, within a week (unless we’re actively gathering information to decide where it belongs). Also note that sometimes meta bugs don’t have a better component, and in these cases, *::General is an appropriate long term home.

  • (Core::General has its own Definitions for Triage Owner Rotations)

Backlogs

Backlogs are a feature of all software projects and we’re never going to get to zero bugs. However, given our already sizable backlogs, we should at least ensure that our backlogs are not getting any worse. We have some tools to track this on https://bugdash.moz.tools/. It has an “Overview” tab that shows you the maintenance trends. Ideally over the past 12 weeks, the “Maint Effect” (i.e. Maintenance Effectiveness) should be over 100%, and the “Burn Down” (i.e. the time to zero bugs) should not be “∞”.

What is a Triaged Bug

The new definition of Triaged will be Firefox-related bugs of type defect where the component is not UNTRIAGED, and a Severity value not equal to -- or N/A.

Bugs of type Task or Enhancement may have a Severity of N/A, but defects must have a Severity that is neither -- nor N/A.

Why Triage

We want to make sure that we looked at every defect and a severity has been defined. This way, we make sure that we did not miss any critical issues during the development and stabilization cycles.

Staying on top of the bugs in your component means:

  • You get ahead of critical regressions and crashes which could trigger a point release if uncaught

    • And you don’t want to spend your holiday with the Release Management team (not that they don’t like you)

  • Your bug queue is not overwhelming

    • Members of your team do not see the bug queue and get the ‘wiggins’

Who Triages

Engineering managers and directors are responsible for naming the individuals responsible for triaging all types of bugs in a component.

We use Bugzilla to manage this. See the list of triage owners.

If you need to change who is responsible for triaging a bug in a component, please file a bug against bugzilla.mozilla.org in the Administration component. When a new component is created, a triage owner must be named.

Rotating triage

Some components are monitored by a rotation of triagers. In those cases, the triage owner on Bugzilla will be automatically updated to reflect the person on the rotation. The rotations are managed as calendars.

If you wish to set up a rotation for triaging one or more components, add a link to your rotation calendar in the triage rotations spreadsheet.

Firefox::General and Toolkit::General

Bugs in Firefox::General are fitted with Bug Bug’s model to see if there’s another component with a high likelihood of fit, and if a threshold confidence is achieved, the bug is moved to that component.

Members of the community also review bugs in this component and try to move them.

What Do You Triage

As a triage owner the queries you should be following for your component are:

  • All open bugs, in your components without a pending needinfo flag which do not have a valid value of severity set

  • All bugs with active review requests in your component which have not been modified in five days

  • All bugs with reviewed, but unlanded patches in your components

  • All bugs with a needinfo request unanswered for more than 10 days

There’s a tool with these queries to help you find bugs https://bugdash.moz.tools/ and the source is at https://github.com/mozilla/bugdash/.

If a bug is an enhancement it needs a priority set and a target release or program milestone. These bugs are normally reviewed by product managers. Enhancements can lead to release notes and QA needed that we also need to know about

If a bug is a task resulting in a changeset, release managers will need to known when this work will be done. A task such as refactoring fragile code can be risky.

Weekly or More Frequently (depending on the component) find un-triaged bugs in the components you triage.

Decide the Severity for each untriaged bug (you can override what’s already been set.)

These bugs are reviewed in the weekly Regression Triage meeting

  • Bugs of type defect with the regression keyword without status-firefoxNN decisions

  • Bugs of type defect with the regression keyword without a regression range

Automatic Bug Updates

When a bug is tracked for a release, i.e. the tracking_firefoxNN flag is set to + or blocking triage decisions will be overridden, or made as follows:

  • If a bug is tracked for or blocking beta, release or ESR, its priority will be set to P1

  • If a bug is tracked for or blocking nightly, its priority will be set to P2

Because bugs can be bumped in priority it’s essential that triage owners review their P1 and P2 bugs frequently.

Assumptions

If a bug’s release status in Firefox version N was affected or wontfix, its Severity is S3 or S4 and its Priority is P3 or lower (backlog,) then its release status in Firefox version N+1, if the bug is still open, is considered to be wontfix.

Questions and Edge Cases

This bug is a feature request

Set the bug’s type to enhancement, add the feature keyword if relevant, and state to NEW. Set the bug’s Severity to N/A. This bug will be excluded from future triage queries.

This bug is a task, not a defect

Set the bug’s type to task, and state to NEW. Set the bug’s Severity to N/A. This bug will be excluded from future triage queries.

If you are not sure of a bug’s type, check our rules for bug types.

This bug’s state is UNCONFIRMED

Are there steps to reproduce? If not, needinfo the person who filed the bug, requesting steps to reproduce. You are not obligated to wait forever for a response, and bugs for which open requests for information go unanswered can be RESOLVED as INCOMPLETE.

I need help reproducing the bug

Set a needinfo for the QA managers, Softvision project managers, or the QA owner of the component of the bug.

I don’t have enough information to make a decision

If you don’t have a reproduction or confirmation, or have questions about how to proceed, needinfo the person who filed the bug, or someone who can answer.

The stalled keyword

The extreme case of not-enough-information is one which cannot be answered with a needinfo request. The reporter has shared all they know about the bug, we are out of strategies to take to resolve it, but the bug should be kept open.

Mark the bug as stalled by adding the stalled keyword to it. The keyword will remove it from the list of bugs to be triaged.

If a patch lands on a stalled bug, automation will remove the keyword. Otherwise, when the keyword is removed, the bug will have its priority reset to -- and the components triage owner notified by automation.

Bugs which remain stalled for long periods of time should be reviewed, and closed if necessary.

Bug is in the wrong Component

If the bug has a Severity of S3, S4, or N/A move the what you think is the correct component, or needinfo the person responsible for the component to ask them.

If the bug has a Severity of S1 or S2 then notify Release Management and contact the triage owner of the component for which you think it belongs to. We cannot lose track of a high severity bug because it is in the wrong component.

My project is on GitHub

We have a guide for GitHub projects to follow when triaging. (Note: this guide needs updating.)

Summary

Multiple times weekly

Use queries for the components you are responsible for in https://github.com/mozilla/bugdash/ to find bugs in need of triage.

For each untriaged bug:

  • Assign a Severity

  • Do not assign a defect a Severity of N/A

You can, but are not required to set the bug’s Priority.

Watch open needinfo flags

Don’t let open needinfo flags linger for more than two weeks.

Close minor bugs with unresponded needinfo flags.

Follow up on needinfo flag requests.

BugDash will help you find these.

End of Iteration/Release Cycle

Any open S1 or S2 bugs at the end of the release cycle will require review by engineering and release management. A policy on this is forthcoming.

Optional

(The guidelines on bug priority are under review.)

Are there open P1s? Revisit their priority, and move to them to the backlog (P3) or P2.

Are there P2 bugs that should move to P1 for the next cycle?

Are there P2 you now know are lower priority, move to P3.

Are there P3 bugs you now know you won’t get to? Either demote to P5 (will accept patch) or resolve as WONTFIX.

Getting help

  • Ask in #bug-handling on chat.mozilla.org