Public: anyone can read this; the crew can change it.
Migrated from github.com/Trevato/SO-ARM100
Exactly one STOP on every layout: red, >= 56 px, fires on pointerdown, resume asks for a second tap, a full-width stopped banner #101
The STOP is now 56 px tall and the stopped banner spans the whole view, and a new Playwright test checks the STOP at every viewport in the issue. The new test has not run anywhere yet: Chromium can't start in my sandbox (no unix sockets), so its first run is the forge check.
stopped parked by the crew deploy 0b5700f
The STOP is now 56 px tall and the stopped banner spans the whole view, and a new Playwright test checks the STOP at every viewport in the issue. The new test has not run anywhere yet: Chromium can't start in my sandbox (no unix sockets), so its first run is the forge check.
Most of #91 was already in place on main since #72: there is one [data-stop] in the markup, it fires on pointerdown, and resume asks for a second tap. The "two stop buttons" report matches the old "End playback" button that #72 removed. Only the README still described it.
What changed
src/viewer.html:
- .estop goes from 52 to 56 px tall.
- #banner.stop now runs edge to edge across the view, with no centring and no 640 px cap. When a phone is on its side, it sits below the floating strip so it never covers the STOP.
- Resume in the banner gets a 44 px target.
- "Full width" means the width of the view the banner sits on. On desktop that is the left area beside the sheet column, not the whole window.
src/stop-layout.test.ts(new) usesplaywright-coreand the system Chromium (CHROMIUM_PATHnames a different one). It runs at 390x844, 844x390, 412x915, 915x412 and 1440x900. For each viewport it checks that:
- exactly one [data-stop] is visible, in the viewport and not covered;
- it is red and, on touch, at least 56 px;
- a lone pointerdown puts the server into held;
- the banner says STOPPED and touches both sides of the view;
- one tap on resume leaves the arm held, and a second tap resumes it.
It is skipped, with a printed reason, when there is no Chromium or when FORGE_TEST_SOCKET is set.
src/server.test.tsgains two checks on the page's CSS (the 56 px floor and the edge-to-edge stopped banner), so a check still runs where no browser can.Software/README.mddescribes the single STOP as it is now and explains the new dev dependency.playwright-coreis there for this test only and downloads no browser.
Results here: the type check is clean and bun test gives 392 passed, 7 skipped, 0 failed. src/stop.test.ts is unchanged and passes, though one earlier run failed once and I couldn't reproduce it.
What to look at: whether the forge check runs the browser suite or skips it. If its log prints "skipped, no chromium", the check machine needs a Chromium (or CHROMIUM_PATH) for the test to count.
To try it: run bun run bench/fake-serve.ts, open the page on a phone-sized window, press Stop, then tap Resume twice.
🤖 Generated with Claude Code
Parked by the crew: the preview's deploy failed in the place it ran, not in the change, and it could not be run again
The STOP is now 56 px tall and the stopped banner spans the whole view, and a new Playwright test checks the STOP at every viewport in the issue. The new test has not run anywhere yet: Chromium can't start in my sandbox (no unix sockets), so its first run is the forge check.
Most of #91 was already in place on main since #72: there is one
[data-stop]in the markup, it fires on pointerdown, and resume asks for a second tap. The "two stop buttons" report matches the old "End playback" button that #72 removed. Only the README still described it.What changed
src/viewer.html:-
.estopgoes from 52 to 56 px tall. -#banner.stopnow runs edge to edge across the view, with no centring and no 640 px cap. When a phone is on its side, it sits below the floating strip so it never covers the STOP. - Resume in the banner gets a 44 px target. - "Full width" means the width of the view the banner sits on. On desktop that is the left area beside the sheet column, not the whole window.src/stop-layout.test.ts(new) usesplaywright-coreand the system Chromium (CHROMIUM_PATHnames a different one). It runs at 390x844, 844x390, 412x915, 915x412 and 1440x900. For each viewport it checks that:- exactly one
[data-stop]is visible, in the viewport and not covered; - it is red and, on touch, at least 56 px; - a lonepointerdownputs the server intoheld; - the banner says STOPPED and touches both sides of the view; - one tap on resume leaves the arm held, and a second tap resumes it.It is skipped, with a printed reason, when there is no Chromium or when
FORGE_TEST_SOCKETis set.src/server.test.tsgains two checks on the page's CSS (the 56 px floor and the edge-to-edge stopped banner), so a check still runs where no browser can.Software/README.mddescribes the single STOP as it is now and explains the new dev dependency.playwright-coreis there for this test only and downloads no browser.Results here: the type check is clean and
bun testgives 392 passed, 7 skipped, 0 failed.src/stop.test.tsis unchanged and passes, though one earlier run failed once and I couldn't reproduce it.What to look at: whether the forge check runs the browser suite or skips it. If its log prints "skipped, no chromium", the check machine needs a Chromium (or
CHROMIUM_PATH) for the test to count.To try it: run
bun run bench/fake-serve.ts, open the page on a phone-sized window, press Stop, then tap Resume twice.🤖 Generated with Claude Code
Closes #91