Skip to main content
Version: 0.1.155 – latest

Bug reports

A bug report is a problem raised against a running application, which an agent then investigates. It is the route from "this is wrong" to a fix, without anyone having to work out first which part of the system is at fault.

Raising one​

Bug reports come from wherever you are looking at the application:

  • from a Test Replay, pinned to the step that went wrong;
  • from a UAT session, where your installation has UAT enabled;
  • pinned to a specific component of a prototype, HTML or target-stack.

Pinning matters. A report attached to what you were looking at carries the context an investigation would otherwise have to reconstruct — which screen, which component, which scenario, which run.

You can report several pinned problems from one session as one submission, rather than filing them one at a time.

Reporting from your own staging site​

When your testers use the application where you deploy it rather than inside Repave, they report with the Repave QA browser extension, for Chrome and Edge.

  1. Your IT admin installs it once: see Installing the QA extension. After that, Install QA extension on Bug Reports adds it, or it is already there.
  2. In the extension's settings, enter your Repave server's address and click Sign in. Chrome asks once to let the extension work on the sites you visit — it needs that to take screenshots — and then you sign in on Repave's own page. Its access to Repave is limited to listing your projects and sending reports.
  3. On the staging site, click the Repave icon (or press Alt+Shift+Q) to open the Repave QA panel. The first time, choose which project the site reports to; it is remembered for the site. The panel shows the build the site runs.
  4. The page works as usual until you turn on Pick issues. Then click what is wrong, say why in What's wrong?, set its Severity (Medium unless you change it) and click Add issue. Picking stays on for the next one; press Esc or turn the switch off to use the page again. Add as many issues as you find, on as many pages as you visit: the panel lists them, and you can edit or delete each one.
  5. Send N issues sends them. Closing the panel keeps them, and the Repave icon shows how many are waiting to be sent.

Any member of the project can send issues, Viewers included.

A send becomes one report: each issue is one of its pins, with its own screenshot, page, element and severity, and the report takes the highest severity among them and the build the site was running. Its investigation starts straight away, as for a report raised inside Repave, and Fix with agent fixes all its issues together. Bug Reports shows that build as Seen on: the commit and how far the main branch has moved since, Build unknown when the site does not say, or not in repo when it was built from commits Repave does not have.

Telling Repave which build staging runs​

A server application tells Repave its build at /.well-known/repave-build.json, and only where you switch it on: set REPAVE_BUILD_INFO=true in your staging deployment's environment, never in production. If your reports show Build unknown, Add build info on Bug Reports has an agent add this to your application; rebuild staging with the switch on afterwards. An application served only as static files cannot have it, and its reports stay Build unknown.

On staging — retest appears beside a fixed report once a staging site runs a build that contains its fix. If the problem is still there, send it again.

What happens next​

An investigation agent picks the report up and works out its scope — whether this is a defect in one feature's implementation, something shared, or a specification problem wearing a bug's clothes. Scope decides who fixes it and where, and getting it wrong is what makes bugs bounce.

The investigation writes up what it found against the report, and you can re-run it. If the first pass missed the point, supply context when you re-run rather than repeating it unchanged — the agent gets what you know, and the second attempt is usually a different answer rather than the same one.

Tracking it​

Bug Reports lists what has been raised and where each one got to. A report that results in a code change is tied to that change, so the trail from the pinned screenshot to the commit stays intact.

Where a fix lands depends on the feature's state — see the status table under Implementing a feature. A bug fixed against a feature that has not merged yet lands on that feature's branch; one fixed after it merged goes through the same route any other change does.

In pull-request merge mode, a fix with no unmerged feature or module branch to land on gets its own pull request. The fix starts from your trunk as it is now and stays on its own branch; nothing is merged until the pull request is. When the fix completes, Create pull request appears at the right of Fix Result. It puts only the fix's commits on a new branch from your trunk, named after the bug (for example repave-platform/fix-legacy-toolbar-icons), writes the title and description, and opens the pull request. The button then becomes View pull request #N, and the pull request is listed on Pull Requests as a UAT bug fix. If the bug is fixed again while the pull request is open, Update pull request puts the new fix on the same pull request. If opening it fails — for example, because the fix no longer applies cleanly to your trunk — the reason shows beside the button.

A ticket in your issue tracker​

When the project has an issue tracker (Settings → Integrations → Issue tracker), every bug report, wherever it was raised, gets one ticket there: a Jira ticket for Jira, a GitHub issue for GitHub Issues (in direct mode too). Nothing for None. Raising a report never waits for the tracker; the ticket appears within a minute.

  • What it says: the report's title, type, severity, who reported it, the build it was seen on, each pinned issue with its page and element, and a link back to the report. A Jira ticket also carries the pins' screenshots, and is a Bug when the Jira project has that type.
  • Investigation: when an investigation completes, Repave adds its summary, suspected root cause and recommended fix as a comment.
  • Status, from Repave: a fix starting moves a Jira ticket to In Progress; a fix completing, or the report marked resolved or won't fix, adds a comment and moves it to Done (closes a GitHub issue). Reopening the report reopens the ticket. Repave picks Jira statuses by their category, so your workflow's own names work; if the workflow allows no move there, the comment is added and the status left.
  • Status, from the tracker: closing a ticket resolves its report (won't fix, for a GitHub issue closed as not planned or a Jira resolution such as Won't Do); reopening it reopens the report. Comments by the project's people appear in the report's discussion, and the agents read them.
  • Ticket on Bug Reports, and the chip on the report, link the ticket. When Repave cannot reach it, the chip says why, and for a token problem what to fix in Settings; saving a new token is all you do.

Asking Repave from a ticket​

Comment on any ticket, one Repave opened or not, starting with /repave and saying what you want:

/repave investigate
/repave fix this, it also happens on the invoice page

Repave reads the tracker every couple of minutes. On a ticket it does not know, it first makes a bug report from the ticket, with its comments and files. Then it starts the investigation or the fix, and replies on the ticket. When it cannot tell which you mean, it asks; when a run is already going or the project's budget is used up, it says so. Anyone who can comment on the ticket can ask; on GitHub, only the repository's owners, members and collaborators. Tickets nobody tagged are never read into Repave.

Next: Jobs.