Repository navigation
feat(view): camera shows what it read — running Scanned list + real shutter (0.17.0) - #35
Conversation
…d a real shutter (0.17.0) - Put away: a Scanned strip right under the video, newest three first with a count, so each read is visibly in without scrolling past the viewfinder. Check a bin shows Found the same way. - A read flashes a quick white shutter, snaps the frame green and names the piece in a pill on the video (plus a short vibrate where supported). Already scanned is amber, a foreign QR is red. - Fix: the old success flash added a bare `hit` class, which is the Stock list's tap-target rule (display:flex + padding), so every read briefly re-laid-out the viewfinder. Classes are cam--*. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…strip; no amber for the label still in view Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…no longer slides in again on every read Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
QA panel review — FAIL
code-review-structural · head 32d2b19bd1a6 · formal
All four findings were confirmed by the verifier with no changes to severity or substance. The one thing to fix before merge is the camHit logic error at view.py:2064: holding an already-added label in view fires a spurious red "Not added" signal every 2.5 s, contradicting both the code comment and the PR description. The two minor regressions (focus-trap escape in the in-place repaint path, and reduced feedback for strays in Check a bin) are worth addressing but are not merge-blockers. No panel disagreement; no coverage gaps.
Findings
| Severity | Location | Finding | Verified | |
|---|---|---|---|---|
| 🟠 | major | view.py:2064 |
In Put away, holding the piece just read in view makes the final else fire a red 'Not added' error signal every dedupe window, although the code comment says… |
confirmed |
| 🟡 | minor | view.py:2451 |
The new in-place paintDrawer path returns before the old focus handling; pressing a button inside the drawer destroys the focused element and leaves focus on… |
confirmed |
| 🟡 | minor | view.py:2053 |
In Check a bin, a freshly read stray or unknown label used to beep and flash; it now gets only the silent amber "dup" signal, which is documented as "Already s… | confirmed |
| ⚪ | nit | pyproject.toml:1 |
Missing [build-system] table in pyproject.toml; pip falls back to setuptools implicitly, which can cause issues in strict or minimal CI environments. | confirmed |
findings JSON (machine-readable)
[
{
"file": "view.py",
"line": 2064,
"severity": "major",
"category": "correctness",
"claim": "In Put away, holding the piece just read in view makes the final `else` fire a red 'Not added' error signal every dedupe window, although the code comment says it should say nothing. Flagged by both correctness and removed-behavior review.",
"evidence": "else if (had && CAM.lastItem !== lab.id) { camSignal(\"dup\", ...); ... CAM.lastItem = lab.id; }\n // had, and it's the piece just read: it's still in view, so say nothing\n else camSignal(\"bad\", PA.note || \"Not added\"); \u2014 when had is true and CAM.lastItem === lab.id the first two branches are skipped and this else runs; paApply has just set PA.note = \"\" so it shows 'Not added'. CAM_DEDUPE_MS = 2500, so a held label re-triggers it every 2.5s.",
"verdict": "confirmed",
"note": "Read view.py:2060-2064 at head: the else branch fires camSignal(\"bad\", PA.note || \"Not added\"). Read paApply at line 2141: for an already-listed item it sets PA.note = \"\" (falsy), so the signal shows 'Not added'. The comment on line 2063 says 'say nothing' and the PR description says 'stays quiet', but the code fires a red error signal. CAM_DEDUPE_MS=2500 (line 1880) gates re-fires.",
"scope": "in-diff"
},
{
"file": "view.py",
"line": 2451,
"severity": "minor",
"category": "removed-behavior",
"claim": "The new in-place `paintDrawer` path returns before the old focus handling; pressing a button inside the drawer destroys the focused element and leaves focus on `<body>`, so keyboard focus can escape the `aria-modal` drawer.",
"evidence": "The old rebuild path ended with `if (!top $(\"#overlay-root .pl-drawer\").focus();`. The new early path (`scroll.innerHTML = body; ... scroll.scrollTop = top; return;`) skips it. The Tab trap only acts when `document.activeElement` is the first or last control, or the box (`document.activeElement === first || document.activeElement === box`). With focus on `<body>`, Tab and Shift+Tab are not intercepted, so focus can move to the page behind the `aria-modal` drawer.",
"verdict": "confirmed",
"note": "Read paintDrawer at line 421: in-place path does scroll.innerHTML = body then return; without refocusing. Read focus trap at line 2451: intercepts only when activeElement is first/last/box. With focus on <body> after innerHTML replacement, neither condition matches and focus escapes the aria-modal drawer.",
"line_original": 436,
"line_corrected": true
},
{
"file": "view.py",
"line": 2053,
"severity": "minor",
"category": "removed-behavior",
"claim": "In Check a bin, a freshly read stray or unknown label used to beep and flash; it now gets only the silent amber \"dup\" signal, which is documented as \"Already scanned\" but here means \"doesn't belong here\".",
"evidence": "Old: `if (bcApply(parseLabel(text), text)) { camBeep(); camFlash(); }`. New: `else if (fresh) camSignal(BC.found.has(lab.id) ? \"ok\" : \"dup\", ...)`. `bcApply` returns `fresh` for the `elsewhere`, `loose`, `sold` and `unknown` kinds, and none of those add to `BC.found`, so they all take the silent \"dup\" path.",
"verdict": "confirmed",
"note": "Diff shows removed line: `if (bcApply(parseLabel(text), text)) { camBeep(); camFlash(); }`. Read bcApply at line 2253: strays (elsewhere/loose/sold) set fresh=true but don't add to BC.found. Read camSignal at line 1934: only kind==='ok' calls camBeep(). So strays get amber visual but no beep \u2014 a reduction from the old beep+flash.",
"line_original": 2051,
"line_corrected": true
},
{
"file": "pyproject.toml",
"line": 1,
"severity": "nit",
"category": "conventions",
"claim": "Missing [build-system] table in pyproject.toml; pip falls back to setuptools implicitly, which can cause issues in strict or minimal CI environments.",
"evidence": "The project has a [project] table but no [build-system] table. Per PEP 518, pip falls back to setuptools when no build-system is declared, so the package will still build. However, this is an implicit default that can cause issues in strict environments (e.g., reproducible-build tooling, some CI images that lack setuptools pre-installed, or package managers like uv that require explicit build-system).",
"source": "protopatch",
"verdict": "confirmed",
"note": "Read pyproject.toml at head: has [project], [tool.ruff], [tool.pytest.ini_options] but no [build-system]. Pre-existing condition (PR only bumps version); valid nit."
}
]…e drawer, strays beep in Check a bin Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
QA panel review — PASS
code-review-structural · head 771b496aab33 · formal
No findings were raised, so there was nothing to verify. The panel confirmed all three round-1 defects (the "Not added" false positive at view.py:2064, the silent stray beep at view.py:2053, and the focus escape at view.py:2451) are fixed in the residual diff. No new defects surfaced. No coverage gaps.
Prior requests
| Prior finding | Disposition | Why | |
|---|---|---|---|
| ✅ | view.py:2064 |
fixed | The residual diff removes the else-branch that fired the red 'Not added' signal when a piece was already in view; the panel confirmed the fix across all five f… |
Findings
| Severity | Location | Finding | Verified | |
|---|---|---|---|---|
| ⚪ | nit | pyproject.toml:1 |
Missing [build-system] table in pyproject.toml; pip falls back to setuptools implicitly, which can cause issues in strict or minimal CI environments. | confirmed |
findings JSON (machine-readable)
[
{
"file": "pyproject.toml",
"line": 1,
"severity": "nit",
"category": "conventions",
"claim": "Missing [build-system] table in pyproject.toml; pip falls back to setuptools implicitly, which can cause issues in strict or minimal CI environments.",
"evidence": "The project has a [project] table but no [build-system] table. Per PEP 518, pip falls back to setuptools when no build-system is declared, so the package will still build. However, this is an implicit default that can cause issues in strict environments (e.g., reproducible-build tooling, some CI images that lack setuptools pre-installed, or package managers like uv that require explicit build-system).",
"source": "protopatch",
"verdict": "confirmed",
"note": "Read pyproject.toml at head: has [project], [tool.ruff], [tool.pytest.ini_options] but no [build-system]. Pre-existing condition (PR only bumps version); valid nit. \u2014 carried unchanged: this file is byte-identical to the round that raised it (incremental, #313)",
"carried": true,
"carried_by": "incremental",
"since": "32d2b19bd1a6d68c6430d2f1314268ff2496a527"
}
]Incremental re-review (#313): 2 of 7 file(s) changed since head 32d2b19bd1a6 (or depend on a change) and were reviewed. 5 unchanged file(s) were not re-read. 1 finding(s) on them are carried forward unchanged.
Comment @vera review for a full re-review.
Addressed since the last round (1 minor/nit): view.py:2053 — no longer raised, and the PR changed the code around them since they were raised.
Superseded — a later head cleared the QA panel (see the newest verdict).
There was a problem hiding this comment.
Promoting the PASS verdict for head 771b496aab33: all checks terminal-green, zero unresolved review threads. (approve-on-green)
Open findings carried by this approval — non-blocking, but they did not go away:
- nit
pyproject.toml:1— Missing [build-system] table in pyproject.toml; pip falls back to setuptools implicitly, which can cause issues in strict or minimal CI environments.
Approving a WARN does not resolve its findings (issue #22).
Josh used Put away for real: it worked, but the camera view needed love — no way to see that things were scanning, and the capture animation was poor.
What changes
+ N more below. Before, the list sat below a 60vh square, i.e. off-screen on a phone while scanning. Check a bin shows Found · x of y the same way. Only a freshly read row slides in; repaints don't replay it.navigator.vibratewhere supported. Already scanned → amber pill/frame; a QR that isn't an Inventory label → red shake.prefers-reduced-motionrespected.hitclass to the viewfinder — which is the Stock list's tap-target rule (.hit{display:flex;padding…}), so every read briefly re-laid-out the camera box. Classes are nowcam--ok/dup/bad.paintDrawerreplaced the whole overlay on every repaint, which replayed the kit's bottom-sheet slide-in on every label read (measured live: the sheet jumped ~500 px and came back). An open drawer on the same side is now updated in place; this helps every drawer, not just Put away.Live check (merchantAgent, branch installed)
Real console through the hub at 390×844, Chromium fake camera playing two real item labels (NH-SH-HARROWS, NH-SH-TORMENT): both read → green + pill + Scanned list; re-shown → amber; 80 samples at 150 ms show the sheet no longer moves. Nothing was moved into a bin; the tray was cleared after.
Tests
487 pass (ruff clean). New: feed ordering/cap/escape/animate-only-new under node; no
hitclass in camera code; all three outcomes signalled; the feed is painted by all three drawers that own the camera.Draft: UI — needs Josh's on-phone check before merge.
🤖 Generated with Claude Code