Skip to content
Visual-diff manual

Start here

Five steps through the deployed console, which runs on committed sample data — nothing you click can start a job or delete anything. Each step links you into the real thing, and names the acceptance scenario that makes the step true.

  1. Open the console

    It opens on what has been photographed so far: a row per screenshot set, carrying the commit and branch behind it and how many stories it holds. This instance is deployed, so every set is a committed sample and nothing you do can change one.

    Checked at merge by “The console lists the screenshot sets”
    1. When I visit the console
    2. Then I see each screenshot set with its branch, story count and size
    3. And a set captured from a dirty tree is marked as dirty
    See the sets

  2. Ask for a comparison

    Pick a before and an after, and the job form fills itself from the choice. There is no button to run it here — a deployed console has no checkout to photograph, and it says so rather than drawing a control that cannot work. This is the shape of the ask, not the running of it.

    Checked at merge by “Comparing two sets pre-fills the job form”
    1. When I visit the console
    2. And I choose two sets to compare
    3. Then the job form is set to compare those two sets
    Set up a comparison

  3. Read the verdict

    A finished comparison opens on its counts — one per bucket, with both sets named above them. Six things moved in this run and a hundred did not, which is the ratio the whole tool exists to make readable.

    Checked at merge by “The report summarizes the comparison by bucket”
    1. When I open the report
    2. Then I see a count chip for every bucket
    3. And both capture sets are identified above the results
    Open a finished report

  4. Look closely at one change

    The comparison opens over the report: before, after, and the diff the run generated, with a slider to move the boundary between them. This link opens it for you — that a link can is a scenario of its own.

    Checked at merge by “A shared link restores the modal state”
    1. When I open a report link carrying a story and slider mode
    2. Then the comparison modal is open with slider mode active
    Open the comparison

  5. Work through the rest

    Marking a variant reviewed advances a count, and the report will jump you to the next one it has not been told about. That loop, not the pixel threshold, is what decides whether a change lands.

    Checked at merge by “Marking a variant reviewed advances the progress”
    1. When I open the report
    2. And I mark the first changed story as reviewed
    3. Then the review progress increases by one
    Review the six changes

That is the loop. The three reference pages carry every scenario behind it, including the ones no tour step needed.