fix(unfinished): fix mobile sizing on Unfinished page and its modals by tstapler · Pull Request #210 · tstapler/stapler-squad

tstapler · 2026-07-23T03:31:22Z

Summary

The Unfinished page and its modal components had no mobile CSS treatment — no dvh/--viewport-height usage, no overflowX: hidden, and hover-only action buttons — unlike sibling pages already fixed in commit 759a1006.

What Changed

UnfinishedTab.css.ts: added overflowX: "hidden" to the page container to stop horizontal scroll at mobile widths.
WorktreeDiffModal.css.ts: maxHeight now uses calc(var(--viewport-height, 100dvh) - 4rem) instead of 100vh, plus a full-screen-sheet variant below the md (768px) breakpoint.
CommitPushModal.tsx: ported to createPortal(document.body), matching WorktreeDiffModal, so its fixed overlay can't be broken by a transformed ancestor.
CommitPushModal.css.ts: overlay/modal now scroll (overflowY: auto) with a dvh-based maxHeight, plus a full-screen-sheet variant below 768px, so Submit/Cancel stay reachable when the on-screen keyboard shrinks the viewport.
UnfinishedItem.css.ts: .actions now forces opacity: 1 under @media (hover: none) so Dismiss/Snooze are visible/tappable on touch devices without requiring hover.
UnfinishedItemDetail.css.ts / WorktreeDiffModal.css.ts: swapped two hardcoded zIndex numbers for the named zIndex.dropdown/zIndex.modal tokens per css-architecture.md.

All new rules are scoped inside @media (max-width: 768px) or @media (hover: none) blocks — the existing desktop (>=1024px) maxWidth: 960px, centering, and flexWrap rules are untouched.

Test plan

npx tsc --noEmit — no errors introduced in touched files
pnpm run lint:css (stylelint) — clean
npx --yes tsx scripts/check-theme-contrast.ts — all WCAG AA pairs pass across all 4 themes
npx jest --no-coverage --testPathPatterns="unfinished" — 234/236 suites pass (2 pre-existing unrelated failures in SessionDetail/BacklogEmptyState, not touched by this change)
Full Playwright e2e (navigation.spec.ts, the only spec hitting /unfinished) — could not run in the dev sandbox (no Chromium binary installed); needs a CI or local run with npx playwright install

Unfinished page and its component tree had no mobile CSS treatment, unlike sibling pages fixed in 759a100. Adds overflowX:hidden to the page container, converts WorktreeDiffModal/CommitPushModal to dvh-based sizing with a full-screen-sheet variant below 768px, ports CommitPushModal to createPortal so its fixed overlay can't be broken by a transformed ancestor, and makes item-card Dismiss/Snooze buttons visible via @media (hover: none) instead of hover-only. Also swaps two hardcoded zIndex values for the named theme-contract tokens. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UaXkDT9P1nSwknvACLd6Gi

github-actions · 2026-07-23T03:34:29Z

✅ Registry Validation

Registry Validation
===================

Building backend scanner...
Scanning backend features...
Wrote 117 feature files to /tmp/tmp.vEIzDX6BXx/backend
Wrote 15 feature files to /tmp/tmp.vEIzDX6BXx/backend
Wrote 45 feature files to /tmp/tmp.vEIzDX6BXx/backend
Wrote 7 feature files to /tmp/tmp.vEIzDX6BXx/backend
Wrote 11 feature files to /tmp/tmp.vEIzDX6BXx/backend

=== Backend Registry Diff ===
Committed: 179  Generated: 179  Divergence: 0.0%
⚠️  111 feature(s) missing // +api: marker (markerFound: false)

✅ Registry validation passed. Divergence: 0.0%

Test Coverage: 18/179 features have testIds (10.1%)

Divergence > 2% blocks merges. Coverage reporting is advisory only.

github-actions · 2026-07-23T03:37:51Z

UX Analysis

Check	Status	Details
✅ Axe Core (WCAG 2.1 AA)	success	Critical/serious violations block merge
⚠️ Lighthouse Performance	Score: unknown	Warning if < 70 (non-blocking)
🤖 Claude UX Analysis	Advisory	See docs/qa/ for findings

Axe Core excludes terminal rendering areas (intentional design).
Lighthouse runs in desktop preset for this developer tool.

github-actions · 2026-07-23T03:40:17Z

Go Benchmarks (Tier 1)

No baseline yet — will be established on next push to main.

github-actions · 2026-07-23T03:41:51Z

Frontend Terminal Throughput

No baseline yet — will be established on next push to main.

github-actions · 2026-07-23T03:43:08Z

🎬 E2E Feature Demos

2 shard(s) recorded feature flows for this PR.

recordings shard 1
recordings shard 2

Demo preview opens directly in browser (single-file HTML). Raw WebM recordings in ZIP. Expires after 30 days.

github-actions · 2026-07-23T03:43:48Z

E2E RPC Latency

No baseline yet — will be established on next push to main.

github-actions · 2026-07-23T03:44:51Z

📊 Feature E2E Coverage

Feature coverage report unavailable

Run make e2e-report locally to view the full Allure report.

The suite couldn't execute at all until now (no Chromium in this sandbox previously masked it): rpc() was missing the /api prefix every RPC route requires, the pinned-repo fixture pointed at the bare repo instead of the working clone (scanner needs a .git dir to resolve `git worktree list`), and the hardcoded session title collided across hook retries against the same long-lived test server. All three predate this branch (authored 2026-04-30, commit 3296f2c). With these fixed, the suite actually runs: the accessibility contrast test and 7/11 core-mechanics tests (page load, empty state, filter toggle, refresh RPC, item expand/collapse, nav badge, reattach session) pass cleanly. The Commit & Push modal test failure confirms the CommitPushModal portal migration works correctly (dialog opens with proper aria-modal/aria-labelledby) — it fails only because its locator is too broad and also matches an unrelated pre-existing Notification Panel dialog. Remaining failures (Open Session fixture mismatch, Dismiss button strict-mode violation from worktree accumulation across sequential beforeAll runs) are pre-existing test/fixture design issues unrelated to the CSS changes in this branch. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UaXkDT9P1nSwknvACLd6Gi

Pre-existing untracked bug docs from prior profiling/investigation work in this shared worktree; committing so they aren't lost. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UaXkDT9P1nSwknvACLd6Gi

github-actions · 2026-07-23T04:02:05Z

✅ Registry Validation

Registry Validation
===================

Building backend scanner...
Scanning backend features...
Wrote 117 feature files to /tmp/tmp.qTZVlQimu5/backend
Wrote 15 feature files to /tmp/tmp.qTZVlQimu5/backend
Wrote 45 feature files to /tmp/tmp.qTZVlQimu5/backend
Wrote 7 feature files to /tmp/tmp.qTZVlQimu5/backend
Wrote 11 feature files to /tmp/tmp.qTZVlQimu5/backend

=== Backend Registry Diff ===
Committed: 179  Generated: 179  Divergence: 0.0%
⚠️  111 feature(s) missing // +api: marker (markerFound: false)

✅ Registry validation passed. Divergence: 0.0%

Test Coverage: 18/179 features have testIds (10.1%)

Divergence > 2% blocks merges. Coverage reporting is advisory only.

…no-op Fixes three pre-existing e2e fixture/locator bugs blocking the suite: the "Open Session" test asserted on a worktree that always had a matching session (so it never showed that button), and the Commit & Push / Dismiss tests used ambiguous role queries that also matched the always-present Notification Panel dialog and the item card's own role="button" wrapper. Also fixes a real scanner bug found while stabilizing the suite: EnqueueRepo's 30s per-worktree TTL cache silently no-ops a manual ScanUnfinishedWork RPC call (the Refresh button) if any worktree in that repo was scanned recently, so repeated clicks did nothing. Periodic ticks should respect the TTL; a user-initiated scan should not. unfinished-work.spec.ts (11/11) and the /unfinished accessibility contrast test now pass consistently across repeat runs. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UaXkDT9P1nSwknvACLd6Gi

…-fix-unfinished-page-mobile-sizing

github-actions · 2026-07-24T06:25:37Z

✅ Registry Validation

Registry Validation
===================

Building backend scanner...
Scanning backend features...
Wrote 117 feature files to /tmp/tmp.breuwWHPkJ/backend
Wrote 15 feature files to /tmp/tmp.breuwWHPkJ/backend
Wrote 45 feature files to /tmp/tmp.breuwWHPkJ/backend
Wrote 7 feature files to /tmp/tmp.breuwWHPkJ/backend
Wrote 11 feature files to /tmp/tmp.breuwWHPkJ/backend

=== Backend Registry Diff ===
Committed: 179  Generated: 179  Divergence: 0.0%
⚠️  111 feature(s) missing // +api: marker (markerFound: false)

✅ Registry validation passed. Divergence: 0.0%

Test Coverage: 18/179 features have testIds (10.1%)

Divergence > 2% blocks merges. Coverage reporting is advisory only.

The Sources-link test added by the main merge (d475144) asserted /settings/unfinished with no trailing slash, but next.config sets trailingSlash: true, so the app always renders the URL with one — the assertion failed on every run regardless of this branch's changes. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UaXkDT9P1nSwknvACLd6Gi

github-actions · 2026-07-24T06:31:44Z

✅ Registry Validation

Registry Validation
===================

Building backend scanner...
Scanning backend features...
Wrote 117 feature files to /tmp/tmp.zNH70CeS7V/backend
Wrote 15 feature files to /tmp/tmp.zNH70CeS7V/backend
Wrote 45 feature files to /tmp/tmp.zNH70CeS7V/backend
Wrote 7 feature files to /tmp/tmp.zNH70CeS7V/backend
Wrote 11 feature files to /tmp/tmp.zNH70CeS7V/backend

=== Backend Registry Diff ===
Committed: 179  Generated: 179  Divergence: 0.0%
⚠️  111 feature(s) missing // +api: marker (markerFound: false)

✅ Registry validation passed. Divergence: 0.0%

Test Coverage: 18/179 features have testIds (10.1%)

Divergence > 2% blocks merges. Coverage reporting is advisory only.

… enqueueRepo The previous fix only bypassed the repo-level TTL gate in enqueueRepo, so a forced (user-triggered) scan got enqueued but scanWorktree still hit its own independent per-worktree cache and returned stale data — the Refresh button appeared to work (a scanCompleted event fired) but the underlying worktree state wasn't actually re-read. force now flows through scanTask -> scanRepo -> scanWorktree, which skips the cache read (and still refreshes it via the existing cache.Set on the OK path). Adds a regression test that fails without this fix. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UaXkDT9P1nSwknvACLd6Gi

Architecture review flagged that removing the TTL-based throttle for forced scans (previous commit) meant a burst of triggers landing close together — fsnotify + manual Refresh + periodic tick — could each enqueue a redundant concurrent scanRepo for the same repo, since nothing tracked "already queued/scanning" independent of the cache. Adds Scanner.inFlight (sync.Map) so enqueueRepo skips re-queuing a repo that's already queued or being scanned; the worker clears the entry once scanRepo/publishResults finish, so the next trigger picks it back up with current state. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UaXkDT9P1nSwknvACLd6Gi

github-actions · 2026-07-24T06:51:14Z

✅ Registry Validation

Registry Validation
===================

Building backend scanner...
Scanning backend features...
Wrote 117 feature files to /tmp/tmp.ZWy3LUdtEp/backend
Wrote 15 feature files to /tmp/tmp.ZWy3LUdtEp/backend
Wrote 45 feature files to /tmp/tmp.ZWy3LUdtEp/backend
Wrote 7 feature files to /tmp/tmp.ZWy3LUdtEp/backend
Wrote 11 feature files to /tmp/tmp.ZWy3LUdtEp/backend

=== Backend Registry Diff ===
Committed: 179  Generated: 179  Divergence: 0.0%
⚠️  111 feature(s) missing // +api: marker (markerFound: false)

✅ Registry validation passed. Divergence: 0.0%

Test Coverage: 18/179 features have testIds (10.1%)

Divergence > 2% blocks merges. Coverage reporting is advisory only.

… /unfinished The Playwright-managed ephemeral test server (tests/e2e/helpers/test-server.ts) never set STAPLER_SQUAD_INSTANCE=e2e-local, so /api/debug/backlog/{seed,mutate}-* (server/server.go) were never registered — in CI or locally. This silently no-op'd every test that depends on them, including the "backlog live updates" accessibility suite (10 tests were failing on createBacklogItemDirect: a JSON parse error from the 404 HTML fallback) and, critically, the "stuck-item chips pass Axe color-contrast on /unfinished" test: with seeding broken, zero stuck items ever rendered, so the Axe scan had nothing to flag — a false-negative that masked two real WCAG AA violations on /unfinished's StuckItemsSection (added by the recent main merge) the whole time. With seeding fixed, Axe correctly flags: - chipActive: primary+textInverse = 4.43:1 (needs 4.5:1) in the clean theme. Fixed by using primaryActive+primaryText instead (6.29:1) — the existing "pressed state" token pair, so this only touches the affected chip, not primary's much broader usage across the app. - stuckReason chip: success+successBg = 3.83:1. Fixed by nudging success from #10b981 to #34d399 (5.06:1) — a one-step-lighter shade of the same green, following this file's existing WCAG-nudge convention. STAPLER_SQUAD_TEST_DIR (set via --test-dir) still takes priority in config.GetConfigDirForDir, so this doesn't change data-directory isolation — only the debug-handler gate and log prefix. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UaXkDT9P1nSwknvACLd6Gi

github-actions · 2026-07-24T07:05:47Z

✅ Registry Validation

Registry Validation
===================

Building backend scanner...
Scanning backend features...
Wrote 117 feature files to /tmp/tmp.aarRW7afJO/backend
Wrote 15 feature files to /tmp/tmp.aarRW7afJO/backend
Wrote 45 feature files to /tmp/tmp.aarRW7afJO/backend
Wrote 7 feature files to /tmp/tmp.aarRW7afJO/backend
Wrote 11 feature files to /tmp/tmp.aarRW7afJO/backend

=== Backend Registry Diff ===
Committed: 179  Generated: 179  Divergence: 0.0%
⚠️  111 feature(s) missing // +api: marker (markerFound: false)

✅ Registry validation passed. Divergence: 0.0%

Test Coverage: 18/179 features have testIds (10.1%)

Divergence > 2% blocks merges. Coverage reporting is advisory only.

…-fix-unfinished-page-mobile-sizing # Conflicts: # tests/e2e/helpers/test-server.ts # web-app/src/components/backlog-stuck/StuckItemsSection.css.ts

github-actions · 2026-07-24T18:58:09Z

✅ Registry Validation

Registry Validation
===================

Building backend scanner...
Scanning backend features...
Wrote 117 feature files to /tmp/tmp.O5QLWQ9c3N/backend
Wrote 15 feature files to /tmp/tmp.O5QLWQ9c3N/backend
Wrote 45 feature files to /tmp/tmp.O5QLWQ9c3N/backend
Wrote 7 feature files to /tmp/tmp.O5QLWQ9c3N/backend
Wrote 11 feature files to /tmp/tmp.O5QLWQ9c3N/backend

=== Backend Registry Diff ===
Committed: 179  Generated: 179  Divergence: 0.0%
⚠️  111 feature(s) missing // +api: marker (markerFound: false)

✅ Registry validation passed. Divergence: 0.0%

Test Coverage: 18/179 features have testIds (10.1%)

Divergence > 2% blocks merges. Coverage reporting is advisory only.

…n backstop Closes the confirmed PR-metadata capture gap (backlog item 12981e9d, PR #210): an agent driving its own shipping via /backlog:ship -> gh pr create never reported the resulting PR back to the item record, since only the system- driven mechanical pushAndCreatePR path wrote pr_url/pr_number. Epic 3.1: new report_pr_created MCP tool (role: work) verifies the reported PR against GitHub (VerifyPRMatchesBranch, reusing GetPRForBranch's typed ErrNoPR) before trusting it, then persists via the new primary-write Storage.SetBacklogItemPRAndTransition and transitions review -> pr_pending. Idempotent on a repeated call for the same PR. Epic 3.2: reconcileOrphanedAgentPRs periodic detector backstops the case where the agent crashes after gh pr create but before calling report_pr_created — finds review-status items with no live session but a real open GitHub PR for their branch, and self-heals via the same SetBacklogItemPRAndTransition path. Deliberately not a new StuckReason per ADR-001/plan.md's scope note (StuckReasonAbandonedReview already covers the genuinely-no-PR case). Frontend verified unchanged: PullRequestSection already renders pr_url/ pr_number correctly whenever item.status === "pr_pending", which both new code paths now reliably reach. project_plans/backlog-agent-communication/implementation/plan.md, Phase 3. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NnUqFpFJid7USqmi7CJQJ7

…n input BUG-047: WriteToSession (web UI chat box) and the write_to_session/ run_command MCP tools appended "\n" after input instead of "\r", so messages sent to a live raw-mode TUI session (the Claude Code CLI itself) were written into the pane but never actually submitted -- they sit unactioned in the input buffer with no error surfaced to the sender. TapEnter, the backlog nudge, the initial-prompt send, and the autonomous driver's turn-submit keystroke all already correctly used "\r" (one with a comment tracing the exact failure mode back to BUG-031). Centralizes the convention into session.EnterKeySequence / session.BuildSubmittableInput and switches all six call sites to it. Found investigating why backlog item 12981e9d (PR #210) sat at status=review for 2.5+ hours: its work session had already gotten the PR fully green and ready to merge, and a human's "merge it" message into the live pane silently failed to submit for exactly this reason. Also files BUG-048 (open, not fixed here): a related but distinct gap where a stuck review-role autonomous session's autonomous_stuck row has no remediation/respawn path, unlike the work-role case. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NnUqFpFJid7USqmi7CJQJ7

… branch Merging PR #210's branch (created before BUG-018-024/027 were closed and moved to docs/bugs/fixed/) reintroduced their pre-resolution copies into docs/bugs/open/ as new files, since that branch's history predated the deletion. The fixed/ versions are authoritative; removing the resurrected open/ duplicates.

tstapler and others added 2 commits July 22, 2026 20:56

tstapler and others added 2 commits July 23, 2026 23:21

Merge remote-tracking branch 'origin/main' into backlog/stapler-squad…

8c70b91

…-fix-unfinished-page-mobile-sizing

tstapler and others added 2 commits July 23, 2026 23:44

Merge remote-tracking branch 'origin/main' into backlog/stapler-squad…

c3ed64a

…-fix-unfinished-page-mobile-sizing # Conflicts: # tests/e2e/helpers/test-server.ts # web-app/src/components/backlog-stuck/StuckItemsSection.css.ts

tstapler merged commit 29eeacf into main Jul 24, 2026
23 checks passed

tstapler deleted the backlog/stapler-squad-fix-unfinished-page-mobile-sizing branch July 24, 2026 21:32

github-actions Bot mentioned this pull request Jul 24, 2026

chore(main): release 1.40.0 #205

Open

tstapler restored the backlog/stapler-squad-fix-unfinished-page-mobile-sizing branch July 24, 2026 22:09

Conversation

tstapler commented Jul 23, 2026

Summary

What Changed

Test plan

Uh oh!

github-actions Bot commented Jul 23, 2026

✅ Registry Validation

Uh oh!

github-actions Bot commented Jul 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

UX Analysis

Uh oh!

github-actions Bot commented Jul 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Go Benchmarks (Tier 1)

Uh oh!

github-actions Bot commented Jul 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Frontend Terminal Throughput

Uh oh!

github-actions Bot commented Jul 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🎬 E2E Feature Demos

Uh oh!

github-actions Bot commented Jul 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

E2E RPC Latency

Uh oh!

github-actions Bot commented Jul 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

📊 Feature E2E Coverage

Uh oh!

github-actions Bot commented Jul 23, 2026

✅ Registry Validation

Uh oh!

github-actions Bot commented Jul 24, 2026

✅ Registry Validation

Uh oh!

github-actions Bot commented Jul 24, 2026

✅ Registry Validation

Uh oh!

github-actions Bot commented Jul 24, 2026

✅ Registry Validation

Uh oh!

github-actions Bot commented Jul 24, 2026

✅ Registry Validation

Uh oh!

github-actions Bot commented Jul 24, 2026

✅ Registry Validation

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

github-actions Bot commented Jul 23, 2026 •

edited

Loading

github-actions Bot commented Jul 23, 2026 •

edited

Loading

github-actions Bot commented Jul 23, 2026 •

edited

Loading

github-actions Bot commented Jul 23, 2026 •

edited

Loading

github-actions Bot commented Jul 23, 2026 •

edited

Loading

github-actions Bot commented Jul 23, 2026 •

edited

Loading