kayg/infraternal
Watch
1
0
Fork
You've already forked infraternal
 0
Code Issues 356 Pull requests Projects Releases Packages Wiki Activity Actions

mobile: text in a terminal window cannot be selected or copied on touch #426

New issue
Open
opened 2026-11-08 04:39:44 +00:00 by kayg · 18 comments
kayg commented 2026-11-08 04:39:44 +00:00
Owner
Copy link

Reported by kayg (2026-11-08), verbatim: "On mobile, I tried to pair T3 code last night and it gave me a pairing code inside infra terminal. But I couldn't select any text at all on mobile. Why is that? Is text selection disabled by touching and dragging inside the terminal pane on mobile? Please debug and fix."

On a phone (iPhone, PWA or Safari), text shown in a terminal window (here, a pairing code printed by a CLI) cannot be selected. Touch and drag in the terminal pane neither starts a selection nor offers copy. Expected: a way to select and copy terminal text on touch devices that feels native (CLAUDE.md: polish on mobile, the PWA must feel native on an iPhone), without breaking scrolling or the existing gestures.

Reported by kayg (2026-11-08), verbatim: "On mobile, I tried to pair T3 code last night and it gave me a pairing code inside infra terminal. But I couldn't select any text at all on mobile. Why is that? Is text selection disabled by touching and dragging inside the terminal pane on mobile? Please debug and fix." On a phone (iPhone, PWA or Safari), text shown in a terminal window (here, a pairing code printed by a CLI) cannot be selected. Touch and drag in the terminal pane neither starts a selection nor offers copy. Expected: a way to select and copy terminal text on touch devices that feels native (CLAUDE.md: polish on mobile, the PWA must feel native on an iPhone), without breaking scrolling or the existing gestures.
kayg commented 2026-11-08 04:41:14 +00:00
Author
Owner
Copy link

J361 start — branch fix-touch-select-426, base f8fe487fe (origin/dev).

Cause (found). The terminal is a Ghostty canvas with no DOM text, so native selection can never apply. Selection is Ghostty's own SelectionManager, driven by mouse events only. A touch drag never produces them, and we also eat the touch:

  • apps/tracer/style.css:174 .terminal-host { touch-action: none } hands every touch to JS.
  • apps/tracer/src/terminal-touch.ts:55 touchmove handler calls event.preventDefault() unconditionally (non-passive), so one finger is always scroll (vertical or horizontal) and two fingers always pinch (D159 "one finger inside a window scrolls"). No long-press path exists.
  • terminal-touch.ts click capture also swallows the click after any move.
    So D159 made one-finger drag = scroll and nothing in the touch model maps to select/copy.

Plan (~3 h): long-press (≈450 ms, <8 px drift) selects the word under the finger (double-tap too); dragging then extends the selection by cell; the shared packages/ui menu shows Copy near the selection; plain swipe still scrolls; desktop mouse path untouched. Spec first (Playwright hasTouch, 393×852 + 375×667, ink+paper), red on origin/dev, then green; iOS Simulator check if the Mac VM allows, otherwise an owner check here.

**J361 start** — branch `fix-touch-select-426`, base `f8fe487fe` (origin/dev). **Cause (found).** The terminal is a Ghostty canvas with no DOM text, so native selection can never apply. Selection is Ghostty's own `SelectionManager`, driven by *mouse* events only. A touch drag never produces them, and we also eat the touch: - `apps/tracer/style.css:174` `.terminal-host { touch-action: none }` hands every touch to JS. - `apps/tracer/src/terminal-touch.ts:55` `touchmove` handler calls `event.preventDefault()` unconditionally (non-passive), so one finger is always scroll (vertical or horizontal) and two fingers always pinch (D159 "one finger inside a window scrolls"). No long-press path exists. - `terminal-touch.ts` click capture also swallows the click after any move. So D159 made one-finger drag = scroll and nothing in the touch model maps to select/copy. **Plan (~3 h):** long-press (≈450 ms, <8 px drift) selects the word under the finger (double-tap too); dragging then extends the selection by cell; the shared `packages/ui` menu shows Copy near the selection; plain swipe still scrolls; desktop mouse path untouched. Spec first (Playwright hasTouch, 393×852 + 375×667, ink+paper), red on origin/dev, then green; iOS Simulator check if the Mac VM allows, otherwise an owner check here.
kayg commented 2026-11-08 06:05:46 +00:00
Author
Owner
Copy link

J361 progress — fix implemented, red/green proven. Branch fix-touch-select-426, head b460d5739, base f8fe487fe.

Cause (line level). apps/tracer/style.css:174 .terminal-host { touch-action: none } hands every touch to JS; apps/tracer/src/terminal-touch.ts touchmove calls event.preventDefault() unconditionally, so one finger is always scroll and two fingers always pinch (D159). The canvas has no DOM text, and Ghostty's own selection listens to mouse events only, which a finger never produces. Nothing in the touch model mapped to select.

Fix.

  • Plain swipe still scrolls (D159 unchanged); two-finger pinch unchanged.
  • A finger resting 450 ms (drift < 6 px) selects the word under it; moving on extends by cell (the first word stays inside, so dragging back past its start keeps it).
  • Double tap selects the word (suited to a pairing code).
  • A Copy callout (the shared openWindowMenu, one row, centred above the selection) copies through writeClipboard. Any later touch of the terminal clears selection and callout.
  • Desktop mouse selection is untouched (new spec guards it).
  • Selection code is *.on-demand.ts, loaded at the first touch (D188); the critical chunk only gains the gesture timer.
  • Touch scroll, pinch and the keyboard open path are unchanged.

Findings worth knowing.

  1. The vendored Terminal.select(col,row,len) anchors rows at the viewport offset, not the scrollback length, so it selects the wrong lines once scrollback exists. I set the selection through the selection manager's own viewportRowToAbsolute instead (typed in one place; pinned by the vendor checksums).
  2. The first run of the green spec exposed a real race: a closing callout's onClosed cleared the reference to the next callout, leaving an orphan menu. Fixed (identity check).
  3. The hooks can still be loading when the finger lifts, so the start is gated on "same touch", not "still selecting".

Evidence (attached below as assets on this comment).

  • Red on spec-only commit 6495e3453: 4/4 fail with Expected "ABCD-EFGH-1234" / Received "" (RED-final.log).
  • Green at 435715969 (same code as head; later commits are comments only): 5/5 pass — 393×852 and 375×667, ink and paper, plus desktop mouse (GREEN-final.log). Covers: long-press word, Copy → clipboard equals the word, tap elsewhere clears, long-press+drag selects second line of output, double tap, plain swipe scrolls and selects nothing.
  • scripts/verify quick on dev-02 at b460d5739: all rows PASS (scope guard, migration names, Nix Cargo vendoring, fmt, doc coverage ratchet, shape lints, doc comments on diff, fixed test waits, ui literals, tracer typecheck).
  • verify fast is queued/running on compute01; result follows in the finish comment.
  • Screenshots: {393x852,375x667}-{ink,paper}-{word,drag}-callout.png.

Not done / owner check. Playwright Chromium (CDP touch) is not iOS Safari. I did not run the iOS Simulator on the Mac VM (that needs the Mac lock and a Simulator setup beyond this slice). Owner check on a real iPhone: open a terminal that shows a pairing code, rest a finger on the code for about half a second (word highlights, Copy appears above it), tap Copy, paste elsewhere; double-tap a word; drag after the long press; swipe to scroll. Please confirm no system magnifier or "Look Up" callout appears over the canvas.

Follow-ups, not in this slice: drag handles to adjust a selection after release (needs a mockup, so not decided silently) and edge auto-scroll while dragging past the top or bottom row.

**J361 progress — fix implemented, red/green proven.** Branch `fix-touch-select-426`, head `b460d5739`, base `f8fe487fe`. **Cause (line level).** `apps/tracer/style.css:174` `.terminal-host { touch-action: none }` hands every touch to JS; `apps/tracer/src/terminal-touch.ts` `touchmove` calls `event.preventDefault()` unconditionally, so one finger is always scroll and two fingers always pinch (D159). The canvas has no DOM text, and Ghostty's own selection listens to mouse events only, which a finger never produces. Nothing in the touch model mapped to select. **Fix.** - Plain swipe still scrolls (D159 unchanged); two-finger pinch unchanged. - A finger resting 450 ms (drift < 6 px) selects the word under it; moving on extends by cell (the first word stays inside, so dragging back past its start keeps it). - Double tap selects the word (suited to a pairing code). - A Copy callout (the shared `openWindowMenu`, one row, centred above the selection) copies through `writeClipboard`. Any later touch of the terminal clears selection and callout. - Desktop mouse selection is untouched (new spec guards it). - Selection code is `*.on-demand.ts`, loaded at the first touch (D188); the critical chunk only gains the gesture timer. - Touch scroll, pinch and the keyboard open path are unchanged. **Findings worth knowing.** 1. The vendored `Terminal.select(col,row,len)` anchors rows at the viewport offset, not the scrollback length, so it selects the wrong lines once scrollback exists. I set the selection through the selection manager's own `viewportRowToAbsolute` instead (typed in one place; pinned by the vendor checksums). 2. The first run of the green spec exposed a real race: a closing callout's `onClosed` cleared the reference to the *next* callout, leaving an orphan menu. Fixed (identity check). 3. The hooks can still be loading when the finger lifts, so the start is gated on "same touch", not "still selecting". **Evidence** (attached below as assets on this comment). - Red on spec-only commit `6495e3453`: 4/4 fail with `Expected "ABCD-EFGH-1234" / Received ""` (`RED-final.log`). - Green at `435715969` (same code as head; later commits are comments only): 5/5 pass — 393×852 and 375×667, ink and paper, plus desktop mouse (`GREEN-final.log`). Covers: long-press word, Copy → clipboard equals the word, tap elsewhere clears, long-press+drag selects `second line of output`, double tap, plain swipe scrolls and selects nothing. - `scripts/verify quick` on dev-02 at `b460d5739`: all rows PASS (scope guard, migration names, Nix Cargo vendoring, fmt, doc coverage ratchet, shape lints, doc comments on diff, fixed test waits, ui literals, tracer typecheck). - `verify fast` is queued/running on compute01; result follows in the finish comment. - Screenshots: `{393x852,375x667}-{ink,paper}-{word,drag}-callout.png`. **Not done / owner check.** Playwright Chromium (CDP touch) is not iOS Safari. I did not run the iOS Simulator on the Mac VM (that needs the Mac lock and a Simulator setup beyond this slice). **Owner check on a real iPhone:** open a terminal that shows a pairing code, rest a finger on the code for about half a second (word highlights, Copy appears above it), tap Copy, paste elsewhere; double-tap a word; drag after the long press; swipe to scroll. Please confirm no system magnifier or "Look Up" callout appears over the canvas. **Follow-ups, not in this slice:** drag handles to adjust a selection after release (needs a mockup, so not decided silently) and edge auto-scroll while dragging past the top or bottom row.
375x667-ink-drag-callout.png
143 KiB
375x667-ink-word-callout.png
136 KiB
375x667-paper-drag-callout.png
127 KiB
375x667-paper-word-callout.png
127 KiB
393x852-ink-drag-callout.png
149 KiB
393x852-ink-word-callout.png
142 KiB
393x852-paper-drag-callout.png
130 KiB
393x852-paper-word-callout.png
130 KiB
RED-final.log
8.4 KiB
GREEN-final.log
3.6 KiB
375x667-ink-drag-callout.png 375x667-ink-word-callout.png 375x667-paper-drag-callout.png 375x667-paper-word-callout.png 393x852-ink-drag-callout.png 393x852-ink-word-callout.png 393x852-paper-drag-callout.png 393x852-paper-word-callout.png
kayg commented 2026-11-08 06:55:28 +00:00
Author
Owner
Copy link

J361 finish — branch fix-touch-select-426, head b460d5739 (base f8fe487fe; 7 commits: test-double hook, red spec, popover extension, the fix, desktop-mouse spec, two doc-comment commits). Not merged, not deployed.

Red → green. Red on spec-only commit 6495e3453: 4/4 fail, Expected "ABCD-EFGH-1234" / Received "". Green at 435715969 (head differs only by comments): 5/5 pass at 393×852 and 375×667, ink and paper, plus a desktop mouse-drag case. Logs and 8 screenshots are attached to the previous comment.

scripts/verify quick (dev-02, head): scope guard, migration names, Nix Cargo vendoring, fmt, doc coverage ratchet, shape lints, doc comments on diff, fixed test waits, ui literals, tracer typecheck — all PASS, ending QUICK pre-check passed.

scripts/verify fast (compute01, head) — did NOT pass; verbatim table:

| Rust gate                 | Result       |
+---------------------------+--------------+
| Nix Cargo vendoring       | PASS         |
| fmt                       | PASS         |
| migration names           | PASS         |
| migration tooling tests   | PASS         |
| VT native fmt             | PASS         |
| CLI plugin compile-out    | PASS         |
| clippy                    | PASS         |
| SSH runtime test helper   | PASS         |
| real-binary fixture paths | PASS         |
| Traefik component manifest | PASS         |
| release component publication | PASS         |
| test                      | FAIL (101)   |
| SSH inode reuse fixture   | PASS         |
| shim process-per-user     | PASS         |
| cargo-deny                | PASS         |
| gate architecture         | PASS         |
| doc coverage ratchet      | PASS         |
| shape lints off (D351)    | PASS         |
| doc comments on diff      | PASS         |
| fixed test waits          | PASS         |
| privileged integration    | SKIP         |
| probe release exclusion   | PASS         |
| nightly fuzz reporter     | PASS         |
| static builds             | PASS         |
+---------------------------+--------------+
FEXIT=1

The only failure is the Rust test row: crates/plugins/ssh/tests/previous_attempt_shell.py, 37 failures such as AssertionError: 73 != 0, next to root-slot-hold: build slot admission failed in the log (compute01 was loaded). My diff contains no Rust. Re-run in isolation, cargo test -p infraternal-plugin-ssh --test previous_attempt_shell: passes on origin/dev 5d34c133a (2 passed, 148 s) and on my head (2 passed, 154 s). So it is a load flake in that row, not this change. The browser rows of fast did not run after the test failure, so the whole tracer Playwright suite has not been run on this branch; only the new spec has (5/5). The orchestrator's merge gate will cover it. Evidence: ~claude/j361-evidence/{fast.log,dev-previous-attempt.log,head-previous-attempt.log} on compute01.

Static size: no Rust or static-binary change, so no D243 delta. The tracer critical chunk gains the gesture timer and double-tap logic in terminal-touch.ts; the selection code and callout are *.on-demand.ts.

Needs the owner: an iPhone check (steps in my previous comment); the iOS Simulator was not run. Selection handles after release and edge auto-scroll are follow-ups, the former needs a mockup.

VM cleanup: target dirs removed on compute01 and dev-02; worktrees ~/worktrees/j361 left.

**J361 finish** — branch `fix-touch-select-426`, head `b460d5739` (base `f8fe487fe`; 7 commits: test-double hook, red spec, popover extension, the fix, desktop-mouse spec, two doc-comment commits). Not merged, not deployed. **Red → green.** Red on spec-only commit `6495e3453`: 4/4 fail, `Expected "ABCD-EFGH-1234" / Received ""`. Green at `435715969` (head differs only by comments): 5/5 pass at 393×852 and 375×667, ink and paper, plus a desktop mouse-drag case. Logs and 8 screenshots are attached to the previous comment. **`scripts/verify quick`** (dev-02, head): scope guard, migration names, Nix Cargo vendoring, fmt, doc coverage ratchet, shape lints, doc comments on diff, fixed test waits, ui literals, tracer typecheck — all PASS, ending `QUICK pre-check passed`. **`scripts/verify fast`** (compute01, head) — did NOT pass; verbatim table: ``` | Rust gate | Result | +---------------------------+--------------+ | Nix Cargo vendoring | PASS | | fmt | PASS | | migration names | PASS | | migration tooling tests | PASS | | VT native fmt | PASS | | CLI plugin compile-out | PASS | | clippy | PASS | | SSH runtime test helper | PASS | | real-binary fixture paths | PASS | | Traefik component manifest | PASS | | release component publication | PASS | | test | FAIL (101) | | SSH inode reuse fixture | PASS | | shim process-per-user | PASS | | cargo-deny | PASS | | gate architecture | PASS | | doc coverage ratchet | PASS | | shape lints off (D351) | PASS | | doc comments on diff | PASS | | fixed test waits | PASS | | privileged integration | SKIP | | probe release exclusion | PASS | | nightly fuzz reporter | PASS | | static builds | PASS | +---------------------------+--------------+ FEXIT=1 ``` The only failure is the Rust `test` row: `crates/plugins/ssh/tests/previous_attempt_shell.py`, 37 failures such as `AssertionError: 73 != 0`, next to `root-slot-hold: build slot admission failed` in the log (compute01 was loaded). My diff contains no Rust. Re-run in isolation, `cargo test -p infraternal-plugin-ssh --test previous_attempt_shell`: **passes on origin/dev `5d34c133a` (2 passed, 148 s) and on my head (2 passed, 154 s)**. So it is a load flake in that row, not this change. The browser rows of `fast` did not run after the `test` failure, so the whole tracer Playwright suite has not been run on this branch; only the new spec has (5/5). The orchestrator's merge gate will cover it. Evidence: `~claude/j361-evidence/{fast.log,dev-previous-attempt.log,head-previous-attempt.log}` on compute01. **Static size:** no Rust or static-binary change, so no D243 delta. The tracer critical chunk gains the gesture timer and double-tap logic in `terminal-touch.ts`; the selection code and callout are `*.on-demand.ts`. **Needs the owner:** an iPhone check (steps in my previous comment); the iOS Simulator was not run. Selection handles after release and edge auto-scroll are follow-ups, the former needs a mockup. VM cleanup: target dirs removed on compute01 and dev-02; worktrees `~/worktrees/j361` left.
kayg commented 2026-11-08 17:04:24 +00:00
Author
Owner
Copy link

J367 review — REQUEST CHANGES · branch fix-touch-select-426 @ b460d5739 (base f8fe487fe) · reviewer J367 (independent UI review; did not merge, deploy or push)

The fix works: long press, double tap, Copy → clipboard, swipe, pinch and the desktop mouse path all behave. I found no P1. Four P2s, mostly against TASTE, mean it should not ship as is.

Findings

P2-1 · The Copy callout overhangs the window and covers its title pill (TASTE §4 "Overlays stay inside the window they belong to").
Measured at 393×852 and 375×667 (393x852-ink.json → s9): for a word on the top row the callout is at top = −7 px relative to the window frame, and for a word at column 0 its left = −14 px. It sits on the window's top border and covers the build-01 title pill (see 393x852-ink-s1-plain-callout.png, 393x852-ink-s9-edge-topleft.png). It does stay ≥ 8 px inside the screen at every edge (min inset 8 px left, 8 px right). The anchor stage is .cl-mockup, not the window. Fix: pass the window as the stage, or clamp to the window box (and flip below the selection when the top row is selected).

P2-2 · Tapping Copy gives no feedback (TASTE "Every tap answers", "Copy without buttons … inline tick plus Copied", "Feedback never depends on success paths you can't see").
run calls writeClipboard and hooks.clear() at once, so the callout simply vanishes. The shared copy animation (tick + "Copied") is not played. On iOS the person cannot tell whether anything was copied. The clipboard is correct ("ABCD-EFGH-1234", no stray newline or space, for both the long press and the double tap).

P2-3 · A word or code that wraps across rows selects only the first row's part, and a drag across the wrap copies a newline.
A 24-char token that wraps at the right edge: long press on its first row selects WRAPTOK, on its second row EN-AAAA-BBBB-CCCC; a drag across the wrap copies "WRAPTOK\nEN-AAAA" (s3* in the JSON). The renderer's own mouse selection gives the identical result (double-click → WRAPTOK), so this is a renderer limit and not a regression in the touch code. But the issue is a pairing code on a 37–39-column phone terminal, and a code or URL longer than the row will not copy intact. Needs the soft-wrapped rows joined (word search across a wrapped row and no \n at a soft wrap), or an explicit owner decision.

P2-4 · The new desktop-mouse spec is flaky; it failed in the full run.
terminal-touch-select.spec.ts:137 failed in the full run, and 1 of 4 with --repeat-each=4 (rerun/select-x4.log). Received "pair: ABCD-EFGH-1234 now\ns" instead of "second line of output": the drag began on row 0. The spec reads the cell centres right after flyView without waiting for the camera to be at rest. It will flake the gate. Wait for the window box to stop moving before taking coordinates.

P3-1 · The selection highlight on ink is hard to read. Selected text is dark on a dim navy fill (393x852-ink-s9-edge-topleft.png, …-s1-plain-callout.png); paper is fine. Not measured against 4.5:1. It is the renderer's selection colours, but on a phone this is the only feedback while the finger is down.
P3-2 · The callout is wide and plain for one word: 168×58 px around one 154×44 row, label 13.5 px / weight 500 (shared menu, contrast 5.3:1 ink and 5.5:1 paper, passes). No pointer to the selection. Hug tighter, and consider whether it should be the icon-with-tooltip form TASTE prefers for copy.
P3-3 · A second finger landing after the hold leaves the selection highlighted with no callout (s11: selection stays ABCD-EFGH-1234, 0 callouts) until the next touch clears it.
P3-4 · A long press inside the host's left padding selects the clamped cell's word (s8 host-left-padding → "second" with a callout). Minor. A long press on the title pill and below the canvas selects nothing.
P3-5 · Not verified on a real iPhone (system magnifier / Look Up / -webkit-touch-callout); that stays the owner check J361 listed. Playwright Chromium CDP touch is not iOS Safari.

What I checked (head b460d5739, production build on dev-02, hasTouch, DPR 3, 393×852 and 375×667, ink and paper)

All four combinations passed my probe (head2/j367/*.json, evidence below).

Check Result
Long press over plain output selects ABCD-EFGH-1234; callout shows
Long press over scrollback (viewportY 29–30) touch = renderer mouse double-click on 3 rows each (tok-085/089/095, tok-097/101/107); drag over 3–4 rows selects the right lines
Across a line wrap see P2-3
Double tap ABCD-EFGH-1234 selects it; Copy → clipboard exactly ABCD-EFGH-1234
Blank space (long press, double tap) no selection, no callout
Swipe scrolls (viewportY 30→49→69), also after a 300 ms rest; selects nothing
Two-finger pinch font 13 → 24; page scale stays 1
Quick tap window active, a textarea inside the host has focus (keyboard raises), no selection
Long press on the title pill no selection, no callout
Callout inside the screen at the edges yes, ≥ 8 px at left/right/bottom of the screen; not inside the window (P2-1)
Legibility 13.5 px; contrast 5.3:1 ink, 5.5:1 paper (pixel-measured)

Popover change (packages/ui): the same probe ran on the base and head builds: row-menu opened at six points (row, near the right edge, near the bottom edge, bottom-right corner, top-left, middle) and the sidebar Account menu. Every box and every inline style (left, top, maxHeight, maxWidth, transformOrigin) is identical (diff empty, cmp-*/j367-compare/*-placement.json). The "More window actions" menu has no ⋯ button in this scene (one action), so it is not covered. By reading: the old flip formula equals the new one for non-above popovers; the only caller of side: "above" (account menu) does not pass at.

Lazy chunk: chunk-psvsf0cm.js, 1,827 B (954 B gzip / 871 B brotli). A desktop session with a mouse selection loaded 88 scripts and not this one; on touch the first tap on the terminal added exactly this chunk (chunk-loading.json). The critical path grows by main.js +89 B (31,093 → 31,182) and +580 B gzip in total (384,441 → 385,021); on-demand total +1.9 KB gzip. transfer-budget.ts fails the critical cap identically on base and head in this environment, so that is not from this branch.

Full tracer Playwright suite (compute01, head, 1,248 tests, one worker, 3.8 h): 997 passed, 147 skipped, 104 failed. The base build is also red here, so I re-ran the failing spec files on base and on head: 92 failures on base, 91 on head. Failing on head and passing on the base re-run: font specs (font-metrics, font-weight, flaky fonts on this host), clusters-keyboard-split (a dropped-frame count; 4/4 pass on both base and head when repeated), and P2-4. Nothing in the touch, terminal-scroll, pinch or popover specs fails on head and not on base. The failures need an orchestrator decision on the red baseline.

Evidence

  • compute01 ~claude/j367-evidence/full/run.log (full suite), rerun/{j367-base,j367-head}.log, rerun/select-x4.log, rerun/ks-*.log
  • dev-02 ~claude/j367-evidence/{head2,cmp-head,cmp-base}/ (JSON + 100+ screenshots); probe specs (untracked, not committed): /tmp/j367/j367-review.spec.ts, j367-compare.spec.ts on the control host
  • Screenshots attached to this comment.

Slots: two (compute01 full suite and dev-02 probes) ran at the same time, since the task asked for both.

**J367 review — REQUEST CHANGES** · branch `fix-touch-select-426` @ `b460d5739` (base `f8fe487fe`) · reviewer J367 (independent UI review; did not merge, deploy or push) The fix works: long press, double tap, Copy → clipboard, swipe, pinch and the desktop mouse path all behave. I found no P1. Four P2s, mostly against TASTE, mean it should not ship as is. ## Findings **P2-1 · The Copy callout overhangs the window and covers its title pill (TASTE §4 "Overlays stay inside the window they belong to").** Measured at 393×852 and 375×667 (`393x852-ink.json` → `s9`): for a word on the top row the callout is at `top` = −7 px relative to the window frame, and for a word at column 0 its `left` = −14 px. It sits on the window's top border and covers the `build-01` title pill (see `393x852-ink-s1-plain-callout.png`, `393x852-ink-s9-edge-topleft.png`). It does stay ≥ 8 px inside the *screen* at every edge (min inset 8 px left, 8 px right). The anchor stage is `.cl-mockup`, not the window. Fix: pass the window as the stage, or clamp to the window box (and flip below the selection when the top row is selected). **P2-2 · Tapping Copy gives no feedback (TASTE "Every tap answers", "Copy without buttons … inline tick plus Copied", "Feedback never depends on success paths you can't see").** `run` calls `writeClipboard` and `hooks.clear()` at once, so the callout simply vanishes. The shared copy animation (tick + "Copied") is not played. On iOS the person cannot tell whether anything was copied. The clipboard is correct (`"ABCD-EFGH-1234"`, no stray newline or space, for both the long press and the double tap). **P2-3 · A word or code that wraps across rows selects only the first row's part, and a drag across the wrap copies a newline.** A 24-char token that wraps at the right edge: long press on its first row selects `WRAPTOK`, on its second row `EN-AAAA-BBBB-CCCC`; a drag across the wrap copies `"WRAPTOK\nEN-AAAA"` (`s3*` in the JSON). The renderer's own mouse selection gives the identical result (double-click → `WRAPTOK`), so this is a renderer limit and not a regression in the touch code. But the issue is a pairing code on a 37–39-column phone terminal, and a code or URL longer than the row will not copy intact. Needs the soft-wrapped rows joined (word search across a wrapped row and no `\n` at a soft wrap), or an explicit owner decision. **P2-4 · The new desktop-mouse spec is flaky; it failed in the full run.** `terminal-touch-select.spec.ts:137` failed in the full run, and 1 of 4 with `--repeat-each=4` (`rerun/select-x4.log`). Received `"pair: ABCD-EFGH-1234 now\ns"` instead of `"second line of output"`: the drag began on row 0. The spec reads the cell centres right after `flyView` without waiting for the camera to be at rest. It will flake the gate. Wait for the window box to stop moving before taking coordinates. **P3-1 · The selection highlight on ink is hard to read.** Selected text is dark on a dim navy fill (`393x852-ink-s9-edge-topleft.png`, `…-s1-plain-callout.png`); paper is fine. Not measured against 4.5:1. It is the renderer's selection colours, but on a phone this is the only feedback while the finger is down. **P3-2 · The callout is wide and plain for one word:** 168×58 px around one 154×44 row, label 13.5 px / weight 500 (shared menu, contrast 5.3:1 ink and 5.5:1 paper, passes). No pointer to the selection. Hug tighter, and consider whether it should be the icon-with-tooltip form TASTE prefers for copy. **P3-3 · A second finger landing after the hold leaves the selection highlighted with no callout** (`s11`: selection stays `ABCD-EFGH-1234`, 0 callouts) until the next touch clears it. **P3-4 · A long press inside the host's left padding selects the clamped cell's word** (`s8` `host-left-padding` → `"second"` with a callout). Minor. A long press on the title pill and below the canvas selects nothing. **P3-5 · Not verified on a real iPhone** (system magnifier / Look Up / `-webkit-touch-callout`); that stays the owner check J361 listed. Playwright Chromium CDP touch is not iOS Safari. ## What I checked (head `b460d5739`, production build on dev-02, `hasTouch`, DPR 3, 393×852 and 375×667, ink and paper) All four combinations passed my probe (`head2/j367/*.json`, evidence below). | Check | Result | |---|---| | Long press over plain output | selects `ABCD-EFGH-1234`; callout shows | | Long press over scrollback (viewportY 29–30) | touch = renderer mouse double-click on 3 rows each (`tok-085/089/095`, `tok-097/101/107`); drag over 3–4 rows selects the right lines | | Across a line wrap | see P2-3 | | Double tap `ABCD-EFGH-1234` | selects it; Copy → clipboard exactly `ABCD-EFGH-1234` | | Blank space (long press, double tap) | no selection, no callout | | Swipe | scrolls (viewportY 30→49→69), also after a 300 ms rest; selects nothing | | Two-finger pinch | font 13 → 24; page scale stays 1 | | Quick tap | window active, a `textarea` inside the host has focus (keyboard raises), no selection | | Long press on the title pill | no selection, no callout | | Callout inside the screen at the edges | yes, ≥ 8 px at left/right/bottom of the screen; not inside the window (P2-1) | | Legibility | 13.5 px; contrast 5.3:1 ink, 5.5:1 paper (pixel-measured) | **Popover change (`packages/ui`):** the same probe ran on the base and head builds: row-menu opened at six points (row, near the right edge, near the bottom edge, bottom-right corner, top-left, middle) and the sidebar Account menu. Every box and every inline style (`left`, `top`, `maxHeight`, `maxWidth`, `transformOrigin`) is identical (`diff` empty, `cmp-*/j367-compare/*-placement.json`). The "More window actions" menu has no ⋯ button in this scene (one action), so it is not covered. By reading: the old `flip` formula equals the new one for non-`above` popovers; the only caller of `side: "above"` (account menu) does not pass `at`. **Lazy chunk:** `chunk-psvsf0cm.js`, 1,827 B (954 B gzip / 871 B brotli). A desktop session with a mouse selection loaded 88 scripts and not this one; on touch the first tap on the terminal added exactly this chunk (`chunk-loading.json`). The critical path grows by `main.js` +89 B (31,093 → 31,182) and +580 B gzip in total (384,441 → 385,021); on-demand total +1.9 KB gzip. `transfer-budget.ts` fails the critical cap identically on base and head in this environment, so that is not from this branch. **Full tracer Playwright suite** (compute01, head, 1,248 tests, one worker, 3.8 h): **997 passed, 147 skipped, 104 failed.** The base build is also red here, so I re-ran the failing spec files on base and on head: **92 failures on base, 91 on head**. Failing on head and passing on the base re-run: font specs (font-metrics, font-weight, flaky fonts on this host), `clusters-keyboard-split` (a dropped-frame count; 4/4 pass on both base and head when repeated), and P2-4. Nothing in the touch, terminal-scroll, pinch or popover specs fails on head and not on base. The failures need an orchestrator decision on the red baseline. ## Evidence - compute01 `~claude/j367-evidence/full/run.log` (full suite), `rerun/{j367-base,j367-head}.log`, `rerun/select-x4.log`, `rerun/ks-*.log` - dev-02 `~claude/j367-evidence/{head2,cmp-head,cmp-base}/` (JSON + 100+ screenshots); probe specs (untracked, not committed): `/tmp/j367/j367-review.spec.ts`, `j367-compare.spec.ts` on the control host - Screenshots attached to this comment. Slots: two (compute01 full suite and dev-02 probes) ran at the same time, since the task asked for both.
375x667-ink-s9-edge-topleft.png
124 KiB
375x667-paper-s9-edge-botright.png
118 KiB
393x852-ink-s1-plain-callout.png
145 KiB
393x852-ink-s2-scrollback-row5.png
294 KiB
393x852-ink-s3-wrap-longpress.png
137 KiB
393x852-ink-s9-edge-topleft.png
133 KiB
393x852-paper-s1-plain-callout.png
131 KiB
popover-base-account.png
562 KiB
popover-base-rowmenu-bottom.png
553 KiB
popover-head-account.png
562 KiB
popover-head-rowmenu-bottom.png
553 KiB
375x667-ink-s9-edge-topleft.png 375x667-paper-s9-edge-botright.png 393x852-ink-s1-plain-callout.png 393x852-ink-s2-scrollback-row5.png 393x852-ink-s3-wrap-longpress.png 393x852-ink-s9-edge-topleft.png 393x852-paper-s1-plain-callout.png popover-base-account.png popover-base-rowmenu-bottom.png popover-head-account.png popover-head-rowmenu-bottom.png
kayg commented 2026-11-08 17:08:29 +00:00
Author
Owner
Copy link

J361 round 2 start — branch fix-touch-select-426, merged origin/dev b31c2da2d. Plan (~4 h, browser specs on dev-02 only): red specs first for P2-1 (callout inside the window, never on the title pill), P2-3 (soft-wrapped word selects whole; no newline at a soft wrap; hard newline kept; touch and mouse, in the shared extraction path), P2-4 (flaky mouse spec waits on a settled window box), then P2-2 (Copy feedback via the shared copy-feedback component), the P3s (ink highlight ≥3:1, hugging callout, second finger keeps the callout, left-padding long press selects nothing). Findings follow as comments.

**J361 round 2 start** — branch `fix-touch-select-426`, merged origin/dev `b31c2da2d`. Plan (~4 h, browser specs on dev-02 only): red specs first for P2-1 (callout inside the window, never on the title pill), P2-3 (soft-wrapped word selects whole; no newline at a soft wrap; hard newline kept; touch and mouse, in the shared extraction path), P2-4 (flaky mouse spec waits on a settled window box), then P2-2 (Copy feedback via the shared copy-feedback component), the P3s (ink highlight ≥3:1, hugging callout, second finger keeps the callout, left-padding long press selects nothing). Findings follow as comments.
kayg commented 2026-11-08 17:18:04 +00:00
Author
Owner
Copy link

J361 round 2 — findings and status (branch fix-touch-select-426, now at a6ad96f19, merged with origin/dev b31c2da2d).

Red first (spec-only commit 7e89633ed, dev-02, production build, log r2-red.log on dev-02 ~claude/j361-evidence/): 11 of 15 browser cases red — 6 wrap cases (touch ×4, mouse ×2: Expected "WRAPTOKEN-AAAA-BBBB-CCCC") and 4 placement/Copy/second-finger/padding cases (callout top/left outside the window body); the unit test fails on the missing export. The 4 original flow cases and the desktop drag stay green, so the new cases fail for their own reasons.

Findings and what I did

  • P2-1 (callout overhang): confirmed. The popover's at mode only knew the screen. New shared options bounds (stay inside the window's content area, one inset in) and avoid + atBelow (an "above" popover that would cover the title pill or the window actions opens below the selection instead). Menus that pass neither place exactly as before.
  • P2-2 (no Copy answer): Copy keeps the row open, the icon turns into the shared tick, the label says "Copied" for 900 ms (the shared copy feedback), then callout and selection go. A failed write says "Couldn't copy". TASTE's in-place text fade cannot apply to canvas text, so the row itself answers.
  • P2-3 (wrap): confirmed, and it is in the renderer's extraction, so I fixed it at the shared path: a new terminal-wrap-selection.ts replaces the selection manager's getSelection and its double-click on the instance. Mouse drag, mouse double-click, the copy on release and both touch gestures all use it. No newline where the lower row is a soft-wrapped continuation; a hard newline still copies as one. The engine knows the wrap flag only for the live screen; for scrolled-off history rows a row counts as continued when the row above is full to its last column (documented; wrong only for a line the program ended at exactly the terminal width).
  • P2-4 (flaky mouse spec): the cause was reading cell centres while the camera still flew. cellCentre now waits until the canvas box is identical for 15 animation frames (an observed signal, no fixed sleep) and every case uses it.
  • P3-1: ink's selected text was #1e1e1e on a 33% accent fill: fill 1.81:1 from the cell, text 1.56:1. selectionColors mixes the accent up until the highlight is 3:1 and the theme's text 4.5:1 (ink: 56%, #4a5e96, 3.06:1 and 5.34:1). Light themes keep their soft highlight (the 3:1 rule is applied to dark themes only, as asked for ink); a unit test checks every theme.
  • P3-2: callout sized to its row (.ui-select-callout drops the shared 10.5 rem minimum).
  • P3-3: a second finger after the hold finishes the selection (Copy shows) and is not a pinch.
  • P3-4: a long press outside the canvas (host padding) selects nothing; a drag past the edge still clamps.

Green x3, verify quick, before/after screenshots and the gate output follow in the finish comment.

**J361 round 2 — findings and status** (branch `fix-touch-select-426`, now at `a6ad96f19`, merged with origin/dev `b31c2da2d`). **Red first** (spec-only commit `7e89633ed`, dev-02, production build, log `r2-red.log` on dev-02 `~claude/j361-evidence/`): 11 of 15 browser cases red — 6 wrap cases (touch ×4, mouse ×2: `Expected "WRAPTOKEN-AAAA-BBBB-CCCC"`) and 4 placement/Copy/second-finger/padding cases (callout `top`/`left` outside the window body); the unit test fails on the missing export. The 4 original flow cases and the desktop drag stay green, so the new cases fail for their own reasons. **Findings and what I did** - **P2-1 (callout overhang):** confirmed. The popover's `at` mode only knew the screen. New shared options `bounds` (stay inside the window's content area, one inset in) and `avoid` + `atBelow` (an "above" popover that would cover the title pill or the window actions opens below the selection instead). Menus that pass neither place exactly as before. - **P2-2 (no Copy answer):** Copy keeps the row open, the icon turns into the shared tick, the label says "Copied" for 900 ms (the shared copy feedback), then callout and selection go. A failed write says "Couldn't copy". TASTE's in-place text fade cannot apply to canvas text, so the row itself answers. - **P2-3 (wrap):** confirmed, and it is in the renderer's extraction, so I fixed it at the shared path: a new `terminal-wrap-selection.ts` replaces the selection manager's `getSelection` and its double-click on the instance. Mouse drag, mouse double-click, the copy on release and both touch gestures all use it. No newline where the lower row is a soft-wrapped continuation; a hard newline still copies as one. The engine knows the wrap flag only for the live screen; for scrolled-off history rows a row counts as continued when the row above is full to its last column (documented; wrong only for a line the program ended at exactly the terminal width). - **P2-4 (flaky mouse spec):** the cause was reading cell centres while the camera still flew. `cellCentre` now waits until the canvas box is identical for 15 animation frames (an observed signal, no fixed sleep) and every case uses it. - **P3-1:** ink's selected text was `#1e1e1e` on a 33% accent fill: fill 1.81:1 from the cell, text 1.56:1. `selectionColors` mixes the accent up until the highlight is 3:1 and the theme's text 4.5:1 (ink: 56%, `#4a5e96`, 3.06:1 and 5.34:1). Light themes keep their soft highlight (the 3:1 rule is applied to dark themes only, as asked for ink); a unit test checks every theme. - **P3-2:** callout sized to its row (`.ui-select-callout` drops the shared 10.5 rem minimum). - **P3-3:** a second finger after the hold finishes the selection (Copy shows) and is not a pinch. - **P3-4:** a long press outside the canvas (host padding) selects nothing; a drag past the edge still clamps. Green x3, `verify quick`, before/after screenshots and the gate output follow in the finish comment.
kayg commented 2026-11-08 17:43:06 +00:00
Author
Owner
Copy link

J361 round 2 finish — branch fix-touch-select-426, head ffa0ab146 (merged origin/dev b31c2da2d; 6 commits after the merge: red specs, popover options, wrap-aware selection, dark-theme selection colours, callout/feedback/padding, second finger). Not merged, not deployed.

Red → green (dev-02, production build, one worker):

  • Red at spec-only commit 7e89633ed: 11 of 15 cases and the unit test failed (log r2-red.log).
  • Green at head, --repeat-each=3: 45 passed (11.6m), 0 failed, i.e. 15 cases × 3 = touch flow, callout placement/Copy answer/second finger/padding at 393×852 and 375×667 in ink and paper, wrap for touch (4) and mouse (2, ink and paper), and the desktop mouse drag (which flaked 1 in 4 before; 3 of 3 now). Unit test selection-colors.test.ts: 20 pass, 0 fail. Logs r2-green-final.log, r2-green.log (an earlier identical x3 run on the previous colour fix, 45 passed).

scripts/verify quick (dev-02, head), verbatim:

| Rust gate                 | Result       |
+---------------------------+--------------+
| scope guard               | PASS         |
| migration names           | PASS         |
| Nix Cargo vendoring       | PASS         |
| fmt                       | PASS         |
| doc coverage ratchet      | PASS         |
| shape lints off (D351)    | PASS         |
| doc comments on diff      | PASS         |
| fixed test waits          | PASS         |
| ui literals               | PASS         |
| tracer typecheck          | PASS         |
+---------------------------+--------------+

QUICK pre-check passed. This is not the gate: run scripts/verify fast before reporting done.

verify fast was not run this round (as instructed; compute01's browser baseline is red).

Screenshots (attached): before = round-1 head (before-*), after = this head (after-*). after-*-edge-callout: the callout for a word on the top row opens below it, inside the window, clear of the build-01 pill; the ink highlight is now #495d95 with the theme's text. after-*-copied: the row says "Copied" (caught mid tick transition). Ghostty's own document click handler clears the highlight as soon as Copy is tapped, so the highlight goes at the tap and the callout a moment later.

Notes for the owner/orchestrator

  • selectionColors now sets white or the background colour as selected text on dark themes where the theme's own text cannot reach 4.5:1 (mocha, rosepine, tokyo, gruvbox, dracula, nord). Light themes are unchanged. If you want the highlight rule on light themes too, say so.
  • History-row wraps are a heuristic (row above full to the last column); the live screen uses the engine's flag. Documented in terminal-wrap-selection.ts.
  • Still owed: the real-iPhone check from my earlier comment (system magnifier / Look Up), now plus: tap Copy and watch the row turn into a tick and "Copied".
**J361 round 2 finish** — branch `fix-touch-select-426`, head `ffa0ab146` (merged origin/dev `b31c2da2d`; 6 commits after the merge: red specs, popover options, wrap-aware selection, dark-theme selection colours, callout/feedback/padding, second finger). Not merged, not deployed. **Red → green** (dev-02, production build, one worker): - Red at spec-only commit `7e89633ed`: 11 of 15 cases and the unit test failed (log `r2-red.log`). - Green at head, `--repeat-each=3`: **45 passed (11.6m), 0 failed**, i.e. 15 cases × 3 = touch flow, callout placement/Copy answer/second finger/padding at 393×852 and 375×667 in ink and paper, wrap for touch (4) and mouse (2, ink and paper), and the desktop mouse drag (which flaked 1 in 4 before; 3 of 3 now). Unit test `selection-colors.test.ts`: **20 pass, 0 fail**. Logs `r2-green-final.log`, `r2-green.log` (an earlier identical x3 run on the previous colour fix, 45 passed). **`scripts/verify quick`** (dev-02, head), verbatim: ``` | Rust gate | Result | +---------------------------+--------------+ | scope guard | PASS | | migration names | PASS | | Nix Cargo vendoring | PASS | | fmt | PASS | | doc coverage ratchet | PASS | | shape lints off (D351) | PASS | | doc comments on diff | PASS | | fixed test waits | PASS | | ui literals | PASS | | tracer typecheck | PASS | +---------------------------+--------------+ QUICK pre-check passed. This is not the gate: run scripts/verify fast before reporting done. ``` `verify fast` was not run this round (as instructed; compute01's browser baseline is red). **Screenshots** (attached): before = round-1 head (`before-*`), after = this head (`after-*`). `after-*-edge-callout`: the callout for a word on the top row opens below it, inside the window, clear of the `build-01` pill; the ink highlight is now `#495d95` with the theme's text. `after-*-copied`: the row says "Copied" (caught mid tick transition). Ghostty's own document click handler clears the highlight as soon as Copy is tapped, so the highlight goes at the tap and the callout a moment later. **Notes for the owner/orchestrator** - `selectionColors` now sets white or the background colour as selected text on dark themes where the theme's own text cannot reach 4.5:1 (mocha, rosepine, tokyo, gruvbox, dracula, nord). Light themes are unchanged. If you want the highlight rule on light themes too, say so. - History-row wraps are a heuristic (row above full to the last column); the live screen uses the engine's flag. Documented in `terminal-wrap-selection.ts`. - Still owed: the real-iPhone check from my earlier comment (system magnifier / Look Up), now plus: tap Copy and watch the row turn into a tick and "Copied".
after-375x667-ink-edge-callout.png
139 KiB
after-375x667-paper-copied.png
134 KiB
after-375x667-paper-drag-callout.png
127 KiB
after-375x667-paper-edge-callout.png
130 KiB
after-393x852-ink-copied.png
149 KiB
after-393x852-ink-edge-callout.png
144 KiB
after-393x852-ink-word-callout.png
144 KiB
after-393x852-paper-edge-callout.png
134 KiB
before-375x667-paper-drag-callout.png
127 KiB
before-393x852-ink-word-callout.png
142 KiB
after-375x667-ink-edge-callout.png after-375x667-paper-copied.png after-375x667-paper-drag-callout.png after-375x667-paper-edge-callout.png after-393x852-ink-copied.png after-393x852-ink-edge-callout.png after-393x852-ink-word-callout.png after-393x852-paper-edge-callout.png before-375x667-paper-drag-callout.png before-393x852-ink-word-callout.png
kayg commented 2026-11-08 18:52:36 +00:00
Author
Owner
Copy link

J388 start — re-review of fix-touch-select-426 @ ffa0ab146 (J361 round 2, comment 86857; answers J367's 86785). Reviewer only: no merge, push or deploy.

Plan: scratch worktree on dev-02 = head merged with origin/dev 9965911c8; production build; touch/wrap/mouse + popover-placement specs at 393×852, 375×667, 1440×900 in ink and paper; verify quick; read the diff for the bounds/avoid/atBelow popover options, the wrap heuristic and the colour rule. Findings follow as separate comments.

**J388 start** — re-review of `fix-touch-select-426` @ `ffa0ab146` (J361 round 2, comment 86857; answers J367's 86785). Reviewer only: no merge, push or deploy. Plan: scratch worktree on dev-02 = head merged with origin/dev `9965911c8`; production build; touch/wrap/mouse + popover-placement specs at 393×852, 375×667, 1440×900 in ink and paper; `verify quick`; read the diff for the `bounds`/`avoid`/`atBelow` popover options, the wrap heuristic and the colour rule. Findings follow as separate comments.
kayg commented 2026-11-08 21:03:45 +00:00
Author
Owner
Copy link

J388 interim — fix-touch-select-426 @ ffa0ab146 merged with origin/dev 9965911c8 (scratch merge 3a816b606, dev-02, production build). Evidence on dev-02: ~claude/j388-run/all.log, ~claude/j388-evidence/head/.

Green so far

  • Author specs (touch select ×8, wrap ×6 incl. mouse ink/paper, desktop mouse drag), popover specs (row-menu, replacement-alignment, account-card, J367 placement probe), 393×852 / 375×667 / 1440×900: all passed (all.log, run A: 45 passed; the 2 failures there were my own probe specs, now fixed).
  • verify quick on the merged scratch worktree: all 10 rows PASS, QUICK_EXIT=0 (all.log, "=== QUICK").
  • Highlight, pixel-measured from real screenshots (head-*-geometry.json): ink #495d95 vs cell 3.02:1, text 5.42:1; paper 1.59:1 / 9.81:1. selectionColors for every theme (colors.ts output): all dark themes ≥ 3.00 highlight and ≥ 4.70 text (nord 4.54 / 4.54, white or bg text as claimed); light themes' text ≥ 4.62.
  • Copy feedback uses the shared confirmAction/swapIcon row (DOM has .swap with is-idle/is-done); the row reads "Copied".
  • Callout inside window and off the pill: the author's spec asserts it at both phone sizes, ink and paper, and passes. My screenshots agree: for a top-row word it opens below, inside the window.
  • Popover options: by reading anchored-popover.ts, the old flip/top/maxWidth formulas reduce to the new ones when bounds/avoid/atBelow/atAlign are absent (at.y + h > bottom ≡ h > roomBelow; leftEdge = screen.left). The J367 placement probe's head output matches J367's base numbers; a fresh base run is queued.

Engine answer (wrap flag for scrollback): the vendored VT engine exports ghostty_terminal_is_row_wrapped for the active screen only; the renderer's own buffer API hard-codes isWrapped = false for scrollback rows (ghostty-web.js ~line 640), and no scrollback wrap export exists in the wasm. So there is no real flag today, and the "row above is full to its last column" heuristic is the best available without a vendor patch. Adding a scrollback_row_wrapped export is a small change to the pinned engine; that is a vendor-patch decision for the owner.

Pending (queued behind a 70-minute job on the single slot): single-row mouse drag differential vs base, history wrap cases, wide-character copy, double-tap on an unfocused terminal. Findings follow.

**J388 interim** — `fix-touch-select-426` @ `ffa0ab146` merged with origin/dev `9965911c8` (scratch merge `3a816b606`, dev-02, production build). Evidence on dev-02: `~claude/j388-run/all.log`, `~claude/j388-evidence/head/`. **Green so far** - Author specs (touch select ×8, wrap ×6 incl. mouse ink/paper, desktop mouse drag), popover specs (`row-menu`, `replacement-alignment`, `account-card`, J367 placement probe), 393×852 / 375×667 / 1440×900: all passed (`all.log`, run A: 45 passed; the 2 failures there were my own probe specs, now fixed). - `verify quick` on the merged scratch worktree: all 10 rows PASS, `QUICK_EXIT=0` (`all.log`, "=== QUICK"). - Highlight, pixel-measured from real screenshots (`head-*-geometry.json`): ink `#495d95` vs cell **3.02:1**, text **5.42:1**; paper 1.59:1 / 9.81:1. `selectionColors` for every theme (`colors.ts` output): all dark themes ≥ 3.00 highlight and ≥ 4.70 text (nord 4.54 / 4.54, white or bg text as claimed); light themes' text ≥ 4.62. - Copy feedback uses the shared `confirmAction`/`swapIcon` row (DOM has `.swap` with `is-idle`/`is-done`); the row reads "Copied". - Callout inside window and off the pill: the author's spec asserts it at both phone sizes, ink and paper, and passes. My screenshots agree: for a top-row word it opens below, inside the window. - Popover options: by reading `anchored-popover.ts`, the old `flip`/`top`/`maxWidth` formulas reduce to the new ones when `bounds`/`avoid`/`atBelow`/`atAlign` are absent (`at.y + h > bottom` ≡ `h > roomBelow`; `leftEdge` = `screen.left`). The J367 placement probe's head output matches J367's base numbers; a fresh base run is queued. **Engine answer (wrap flag for scrollback):** the vendored VT engine exports `ghostty_terminal_is_row_wrapped` for the *active screen only*; the renderer's own buffer API hard-codes `isWrapped = false` for scrollback rows (`ghostty-web.js` ~line 640), and no scrollback wrap export exists in the wasm. So there is no real flag today, and the "row above is full to its last column" heuristic is the best available without a vendor patch. Adding a `scrollback_row_wrapped` export is a small change to the pinned engine; that is a vendor-patch decision for the owner. **Pending (queued behind a 70-minute job on the single slot):** single-row mouse drag differential vs base, history wrap cases, wide-character copy, double-tap on an unfocused terminal. Findings follow.
kayg commented 2026-11-08 22:37:47 +00:00
Author
Owner
Copy link

J388 finding P2-1 · text behind the Copy callout stays legible (TASTE §5 "Glass shows colour, never legible text", "at every transparency level")
At 393×852 the callout floats over dense terminal output and the rows behind it read through the glass: in ink "second line of output" / "third" (393x852-ink-word-callout.png, 393x852-ink-edge-callout.png), "row32…row35 word" (head-393x852-ink-last-row-word.png), and in paper "row1 word1", "row4 word4" through the "Copied" row (head-393x852-paper-copied.png). This is the shared window-menu material; until now it sat over chrome, never over dense text. The owner has asked for this pass before: "ghost text again, please do a pass for the cursed text".
Fix: a callout-specific denser tint or a blur of about 1.5× the line height (≥ 27 px at this size) on .ui-select-callout, tuned in packages/ui. Check with a 2× crop over a dense screen in ink and paper.
Evidence: dev-02 ~claude/j388-evidence/head/j388/.

**J388 finding P2-1 · text behind the Copy callout stays legible (TASTE §5 "Glass shows colour, never legible text", "at every transparency level")** At 393×852 the callout floats over dense terminal output and the rows behind it read through the glass: in ink "second line of output" / "third" (`393x852-ink-word-callout.png`, `393x852-ink-edge-callout.png`), "row32…row35 word" (`head-393x852-ink-last-row-word.png`), and in paper "row1 word1", "row4 word4" through the "Copied" row (`head-393x852-paper-copied.png`). This is the shared window-menu material; until now it sat over chrome, never over dense text. The owner has asked for this pass before: "ghost text again, please do a pass for the cursed text". Fix: a callout-specific denser tint or a blur of about 1.5× the line height (≥ 27 px at this size) on `.ui-select-callout`, tuned in `packages/ui`. Check with a 2× crop over a dense screen in ink and paper. Evidence: dev-02 `~claude/j388-evidence/head/j388/`.
393x852-ink-word-callout.png
144 KiB
393x852-ink-edge-callout.png
142 KiB
head-393x852-ink-last-row-word.png
326 KiB
head-393x852-paper-copied.png
302 KiB
393x852-ink-word-callout.png 393x852-ink-edge-callout.png head-393x852-ink-last-row-word.png head-393x852-paper-copied.png
kayg commented 2026-11-08 22:37:47 +00:00
Author
Owner
Copy link

J388 finding P3-1 · tapping Copy takes focus from the terminal
In head-393x852-paper-copied.png (and the author's 393x852-ink-copied.png) the terminal cursor has turned from a solid bar (head-393x852-ink-last-row-word.png, before the tap) into a hollow underline, and the key row below is sliding out (TASTE §4: the key row exists only while a terminal has input focus). The Copy button is a real <button> that receives focus on tap. On an iPhone with the keyboard up, Copy would dismiss it, which the person did not ask for.
Fix: keep focus where it was: pointerdown/mousedown preventDefault() on the callout row (or focus the terminal's field again after the write). Needs the real-iPhone check on the keyboard.

J388 finding P3-2 · a double tap on an unfocused terminal raises the keyboard on its first tap
head-doubletap.json: before, document.activeElement = BODY; after the first tap, a TEXTAREA inside the host has focus (window active); the second tap then selects the word. On a phone the first tap already opens the keyboard, which resizes the visual viewport under the selection and its callout. By design for a single tap (D191), but a double tap meant to select does not want it. Option: defer the focus tap by DOUBLE_TAP_MS (350 ms), or accept and note it. Owner's real-iPhone check.

**J388 finding P3-1 · tapping Copy takes focus from the terminal** In `head-393x852-paper-copied.png` (and the author's `393x852-ink-copied.png`) the terminal cursor has turned from a solid bar (`head-393x852-ink-last-row-word.png`, before the tap) into a hollow underline, and the key row below is sliding out (TASTE §4: the key row exists only while a terminal has input focus). The Copy button is a real `<button>` that receives focus on tap. On an iPhone with the keyboard up, Copy would dismiss it, which the person did not ask for. Fix: keep focus where it was: `pointerdown`/`mousedown` `preventDefault()` on the callout row (or focus the terminal's field again after the write). Needs the real-iPhone check on the keyboard. **J388 finding P3-2 · a double tap on an unfocused terminal raises the keyboard on its first tap** `head-doubletap.json`: before, `document.activeElement` = BODY; after the first tap, a `TEXTAREA` inside the host has focus (window active); the second tap then selects the word. On a phone the first tap already opens the keyboard, which resizes the visual viewport under the selection and its callout. By design for a single tap (D191), but a double tap meant to select does not want it. Option: defer the focus tap by `DOUBLE_TAP_MS` (350 ms), or accept and note it. Owner's real-iPhone check.
393x852-ink-copied.png
149 KiB
head-doubletap.png
134 KiB
393x852-ink-copied.png head-doubletap.png
kayg commented 2026-11-08 22:37:47 +00:00
Author
Owner
Copy link

J388 finding P3-3 · the scrollback wrap heuristic mis-joins or mis-splits (answers the question about a real wrap flag)
There is no real flag for scrollback in this engine build: the wasm exports ghostty_terminal_is_row_wrapped for the active screen only, and the renderer's own buffer API hard-codes isWrapped = false for history rows (ghostty-web.js ~l. 640). Getting one needs an additive export in the pinned engine's vendor patches (owner's decision; Ghostty stores the flag per row internally).
Measured on the heuristic (head-history.json, 393×852, 39 columns; the same screens live vs in history):

  • a program line exactly as wide as the terminal (a full-width rule) followed by after-rule: live screen copies ====…====\nafter-r (correct, engine flag); in history it copies ====…====after-r, newline lost.
  • a soft wrap whose last column holds a blank spacer, because a wide character moved to the next row (CJK text): in history it copies with a \n at the wrap; live it copies xxx… 日 本, a stray space at the wrap.
  • the wrapped pairing-shaped token in history selects and copies whole (WRAPTOKEN-AAAA-BBBB-CCCC): the owner's case works.
    Wide characters copy with a space between them (日 本 語) on base as well (renderer behaviour, not a regression). Acceptable as documented; the loss only happens for lines that scrolled off. Not blocking.
**J388 finding P3-3 · the scrollback wrap heuristic mis-joins or mis-splits (answers the question about a real wrap flag)** There is no real flag for scrollback in this engine build: the wasm exports `ghostty_terminal_is_row_wrapped` for the active screen only, and the renderer's own buffer API hard-codes `isWrapped = false` for history rows (`ghostty-web.js` ~l. 640). Getting one needs an additive export in the pinned engine's vendor patches (owner's decision; Ghostty stores the flag per row internally). Measured on the heuristic (`head-history.json`, 393×852, 39 columns; the same screens live vs in history): - a program line exactly as wide as the terminal (a full-width rule) followed by `after-rule`: live screen copies `====…====\nafter-r` (correct, engine flag); in history it copies `====…====after-r`, newline lost. - a soft wrap whose last column holds a blank spacer, because a wide character moved to the next row (CJK text): in history it copies with a `\n` at the wrap; live it copies `xxx… 日 本`, a stray space at the wrap. - the wrapped pairing-shaped token in history selects and copies whole (`WRAPTOKEN-AAAA-BBBB-CCCC`): the owner's case works. Wide characters copy with a space between them (`日 本 語`) on base as well (renderer behaviour, not a regression). Acceptable as documented; the loss only happens for lines that scrolled off. Not blocking.
kayg commented 2026-11-08 22:37:47 +00:00
Author
Owner
Copy link

J388 note · residual flake on the desktop mouse spec's pattern, not a regression
In my own probe (j388-review.spec.ts, dev-02 ~claude/j388-evidence/{base,head}/) the first single-row drag after a show() mis-selects rows pair: ABCD-EFGH-1234 now\ns instead of the row dragged, identically on base 9965911c8 and on head, and not in the author's spec (passed 3 of 3 for J361, and in my run A). The drag helper's "canvas still for 15 frames" check can pass before a flight begins. The branch did not cause it; the settle wait fixed the author's case but is not proof against this pattern. P3, test-side only.

**J388 note · residual flake on the desktop mouse spec's pattern, not a regression** In my own probe (`j388-review.spec.ts`, dev-02 `~claude/j388-evidence/{base,head}/`) the first single-row drag after a `show()` mis-selects rows `pair: ABCD-EFGH-1234 now\ns` instead of the row dragged, identically on base `9965911c8` and on head, and not in the author's spec (passed 3 of 3 for J361, and in my run A). The drag helper's "canvas still for 15 frames" check can pass before a flight begins. The branch did not cause it; the settle wait fixed the author's case but is not proof against this pattern. P3, test-side only.
kayg commented 2026-11-08 22:38:07 +00:00
Author
Owner
Copy link

J388 final — REQUEST CHANGES (one P2, narrow) · reviewed ffa0ab146, merged with origin/dev 9965911c8 (scratch merge 3a816b606), production build on dev-02, 393×852, 375×667, 1440×900, ink and paper. Did not merge, push or deploy.

J367's findings, each checked

J367 Result
P2-1 callout overhangs window / covers pill Fixed. 7 selections (top-left word, code, mid row, bottom row, row 1, two-row drag) × 4 size/theme combos: callout inside the window frame and content area, never over the title pill or actions (*-geometry.json; spec passes). Below the selection on the top row, above elsewhere.
P2-2 no Copy feedback Fixed. Shared confirmAction/swapIcon (.swap idle→done, label "Copied").
P2-3 wrapped code copies intact Fixed for live screen and history for words/tokens (mouse and touch specs, 6 wrap specs pass); hard newline stays on the live screen. History limits: P3-3.
P2-4 flaky mouse spec Fixed for the author's spec (passes in my runs); my own variant still shows an identical symptom on base too (note, test-side).
P3-1 ink contrast Fixed, pixel-measured: highlight 3.02:1, text 5.42:1 (ink); all dark themes ≥ 3.0 / ≥ 4.54.
P3-2/-3/-4 Callout 92×58 px, 13.5 px label; second finger and padding specs pass.

Popover options: bounds/avoid/atBelow/atAlign are inert when absent; by reading, old and new flip/top/maxWidth agree. Placement probe (6 right-click points, account menu) ran on base and head; specs row-menu, replacement-alignment, account-card pass. No placement regression found.

Desktop mouse: author specs pass at 1440×900 (drag, double-click, wrapped words, hard newline). My differential (base-*-mouse.json vs head-*-mouse.json): double-click words/URL/dashes/blank, triple-click, multi-row drags identical to base; wide-character spaces are baseline behaviour. No regression found.

Findings

  • P2-1 text legible through the Copy callout (comment 87789). Blocks.
  • P3-1 Copy tap blurs the terminal and dismisses the key row (87790).
  • P3-2 double tap on an unfocused terminal raises the keyboard (87790).
  • P3-3 scrollback wrap heuristic limits; no real wrap flag for history in this engine build (87791).
  • Note: residual flake pattern, test-side, not a regression (87792).

Gates

  • Specs run: terminal-touch-select (8 + mouse), terminal-selection-wrap (6), row-menu (5), replacement-alignment (4), account-card (17), J367 placement/tap-focus probes: 45 passed in run A; my probes 8/8 on head, base probes 3/3 (~claude/j388-run/all.log, r2.log).
  • verify quick on the merged scratch worktree: scope guard, migration names, Nix vendoring, fmt, doc coverage ratchet, shape lints, doc comments on diff, fixed test waits, ui literals, tracer typecheck all PASS; QUICK_EXIT=0 (all.log).
  • verify fast not run (reviewer rule).
  • Not covered: real iPhone (system magnifier, Look Up, keyboard behaviour on Copy and double tap).

Verdict: REQUEST CHANGES — reviewed SHA ffa0ab146; gates: specs green, verify quick PASS.

**J388 final — REQUEST CHANGES (one P2, narrow)** · reviewed `ffa0ab146`, merged with origin/dev `9965911c8` (scratch merge `3a816b606`), production build on dev-02, 393×852, 375×667, 1440×900, ink and paper. Did not merge, push or deploy. **J367's findings, each checked** | J367 | Result | |---|---| | P2-1 callout overhangs window / covers pill | **Fixed.** 7 selections (top-left word, code, mid row, bottom row, row 1, two-row drag) × 4 size/theme combos: callout inside the window frame and content area, never over the title pill or actions (`*-geometry.json`; spec passes). Below the selection on the top row, above elsewhere. | | P2-2 no Copy feedback | **Fixed.** Shared `confirmAction`/`swapIcon` (`.swap` idle→done, label "Copied"). | | P2-3 wrapped code copies intact | **Fixed** for live screen and history for words/tokens (mouse and touch specs, 6 wrap specs pass); hard newline stays on the live screen. History limits: P3-3. | | P2-4 flaky mouse spec | **Fixed** for the author's spec (passes in my runs); my own variant still shows an identical symptom on base too (note, test-side). | | P3-1 ink contrast | **Fixed**, pixel-measured: highlight 3.02:1, text 5.42:1 (ink); all dark themes ≥ 3.0 / ≥ 4.54. | | P3-2/-3/-4 | Callout 92×58 px, 13.5 px label; second finger and padding specs pass. | **Popover options:** `bounds`/`avoid`/`atBelow`/`atAlign` are inert when absent; by reading, old and new `flip`/`top`/`maxWidth` agree. Placement probe (6 right-click points, account menu) ran on base and head; specs `row-menu`, `replacement-alignment`, `account-card` pass. No placement regression found. **Desktop mouse:** author specs pass at 1440×900 (drag, double-click, wrapped words, hard newline). My differential (`base-*-mouse.json` vs `head-*-mouse.json`): double-click words/URL/dashes/blank, triple-click, multi-row drags identical to base; wide-character spaces are baseline behaviour. No regression found. **Findings** - **P2-1** text legible through the Copy callout (comment 87789). Blocks. - P3-1 Copy tap blurs the terminal and dismisses the key row (87790). - P3-2 double tap on an unfocused terminal raises the keyboard (87790). - P3-3 scrollback wrap heuristic limits; no real wrap flag for history in this engine build (87791). - Note: residual flake pattern, test-side, not a regression (87792). **Gates** - Specs run: `terminal-touch-select` (8 + mouse), `terminal-selection-wrap` (6), `row-menu` (5), `replacement-alignment` (4), `account-card` (17), J367 placement/tap-focus probes: 45 passed in run A; my probes 8/8 on head, base probes 3/3 (`~claude/j388-run/all.log`, `r2.log`). - `verify quick` on the merged scratch worktree: scope guard, migration names, Nix vendoring, fmt, doc coverage ratchet, shape lints, doc comments on diff, fixed test waits, ui literals, tracer typecheck all PASS; `QUICK_EXIT=0` (`all.log`). - `verify fast` not run (reviewer rule). - Not covered: real iPhone (system magnifier, Look Up, keyboard behaviour on Copy and double tap). **Verdict: REQUEST CHANGES — reviewed SHA `ffa0ab146`; gates: specs green, `verify quick` PASS.**
head-ink-mouse.png
123 KiB
head-paper-mouse.png
120 KiB
375x667-paper-edge-callout.png
131 KiB
393x852-paper-edge-callout.png
134 KiB
375x667-ink-edge-callout.png
139 KiB
head-ink-mouse.png head-paper-mouse.png 375x667-paper-edge-callout.png 393x852-paper-edge-callout.png 375x667-ink-edge-callout.png
kayg commented 2026-11-08 23:17:51 +00:00
Author
Owner
Copy link

J361 — status and a blocker (last small round). Merged origin/dev 0a5432d2c; branch fix-touch-select-426 at 50013d4cf.

Done and pushed: red specs 619e5206a (focus after Copy; no text readable behind the callout: padding-strip detail < 2/255 over a dense screen, plus the glass bed present); then one commit each: the Copy callout on the shared glass bed (glass-bed.ts through a new glassBed option on the shared popover, fixed in the page above the bed) and Copy no longer takes focus (pointerdown/mousedown on the callout cancelled, the click still fires). The keyboard-on-first-tap (double tap) and the scrollback wrap heuristic are left as they are, as instructed.

Blocked: no browser result yet. dev-02's build-slot queue has not moved for over 30 minutes: build-slot --status shows slot 1 held by a fuzz job (2 h 40 min), slots 2 and 3 free, and 7 jobs queued behind a head ticket (pid 471771, a verify quick of j287, 2 h old) that is still alive, so FIFO admission lets nobody take the free slots. My red run is queued at position 7 (~claude/j361-evidence/r3-red.log on dev-02). I did not kill anything that is not mine. I keep polling; if it stays stuck I will report the head ticket to the orchestrator for the Maintainer.

**J361 — status and a blocker (last small round).** Merged origin/dev `0a5432d2c`; branch `fix-touch-select-426` at `50013d4cf`. Done and pushed: red specs `619e5206a` (focus after Copy; no text readable behind the callout: padding-strip detail < 2/255 over a dense screen, plus the glass bed present); then one commit each: the Copy callout on the shared glass bed (`glass-bed.ts` through a new `glassBed` option on the shared popover, fixed in the page above the bed) and Copy no longer takes focus (pointerdown/mousedown on the callout cancelled, the click still fires). The keyboard-on-first-tap (double tap) and the scrollback wrap heuristic are left as they are, as instructed. **Blocked: no browser result yet.** dev-02's build-slot queue has not moved for over 30 minutes: `build-slot --status` shows slot 1 held by a fuzz job (2 h 40 min), slots 2 and 3 free, and 7 jobs queued behind a head ticket (pid 471771, a `verify quick` of j287, 2 h old) that is still alive, so FIFO admission lets nobody take the free slots. My red run is queued at position 7 (`~claude/j361-evidence/r3-red.log` on dev-02). I did not kill anything that is not mine. I keep polling; if it stays stuck I will report the head ticket to the orchestrator for the Maintainer.
kayg commented 2026-11-09 02:57:14 +00:00
Author
Owner
Copy link

J361 last round — finish. Branch fix-touch-select-426, head 3a3466f76 (merged origin/dev 0a5432d2c; three commits on top: red specs e03106d68, glass bed, Copy keeps focus). Not merged, not deployed.

P2 — text behind the Copy callout. The shared popover gains glassBed: it attaches the existing glass-bed.ts under the callout (blur starts well outside the card, using the material's own per-level blur token), the popover becomes position: fixed in the page above the bed, and the bed fades out and goes with it. No one-off tint or blur. Screenshots (attached, before-* vs after-*, dense text screen, ink and paper): before, the row behind the callout reads through its lower and right edges ("row11 lorem ip…"); after, the glass is clean. A pixel guard measures four 3 px strips just inside each edge: before the fix the right edge varied 0.80 (ink) and 0.59 (paper) luminance steps; after, every edge is at most 0.05. The spec requires < 2 on all four edges and the bed element. Honest note: before the fix those strips were already under the 2/255 threshold in this layout, so the red for the glass case comes from the missing bed (the mechanism the material is built on), not from the pixel numbers; the pixels are a guard against a future regression.

P3 — focus. pointerdown and mousedown on the callout are cancelled, so tapping Copy leaves focus in the terminal's input (cursor stays solid, key row stays); the click still fires. Spec: the terminal's input has focus before the selection and after the tap. Needs the iPhone check with the keyboard up.

Left as asked: the double-tap raises the keyboard on its first tap (D191, no added tap latency); the scrollback wrap heuristic (goes to the owner in the morning).

Red → green (dev-02, production build)

  • Red at e03106d68 (spec only): all 8 new cases failed (focus ×4: "tapping Copy leaves focus in the terminal"; glass ×4: no glass bed), log r3-red2.log.
  • Green at head, all touch + wrap + mouse specs, --repeat-each=3: 69 passed (16.3m), 0 failed, plus the colour unit test 20/20. Log r3-green.log. (Two helper bugs in the new spec were fixed on the way: a wait on document.getAnimations() that never resolves because some animations are paused, and a strip that sat on the callout's rounded corner; both are in the spec commit, not the product.)
  • verify quick, verbatim:
| scope guard               | PASS         |
| migration names           | PASS         |
| Nix Cargo vendoring       | PASS         |
| fmt                       | PASS         |
| doc coverage ratchet      | PASS         |
| shape lints off (D351)    | PASS         |
| doc comments on diff      | PASS         |
| fixed test waits          | PASS         |
| ui literals               | PASS         |
| tracer typecheck          | PASS         |
QUICK pre-check passed. This is not the gate: run scripts/verify fast before reporting done.

verify fast not run (as instructed). dev-02 target removed; my two test servers were stopped by PID.

Infra note: dev-02's build-slot queue sat stuck for about 30 minutes earlier (a head ticket that was running, slots 2 and 3 free); it cleared by itself.

**J361 last round — finish.** Branch `fix-touch-select-426`, head `3a3466f76` (merged origin/dev `0a5432d2c`; three commits on top: red specs `e03106d68`, glass bed, Copy keeps focus). Not merged, not deployed. **P2 — text behind the Copy callout.** The shared popover gains `glassBed`: it attaches the existing `glass-bed.ts` under the callout (blur starts well outside the card, using the material's own per-level blur token), the popover becomes `position: fixed` in the page above the bed, and the bed fades out and goes with it. No one-off tint or blur. Screenshots (attached, `before-*` vs `after-*`, dense text screen, ink and paper): before, the row behind the callout reads through its lower and right edges ("row11 lorem ip…"); after, the glass is clean. A pixel guard measures four 3 px strips just inside each edge: before the fix the right edge varied 0.80 (ink) and 0.59 (paper) luminance steps; after, every edge is at most 0.05. The spec requires < 2 on all four edges and the bed element. Honest note: before the fix those strips were already under the 2/255 threshold in this layout, so the red for the glass case comes from the missing bed (the mechanism the material is built on), not from the pixel numbers; the pixels are a guard against a future regression. **P3 — focus.** `pointerdown` and `mousedown` on the callout are cancelled, so tapping Copy leaves focus in the terminal's input (cursor stays solid, key row stays); the click still fires. Spec: the terminal's input has focus before the selection and after the tap. Needs the iPhone check with the keyboard up. **Left as asked:** the double-tap raises the keyboard on its first tap (D191, no added tap latency); the scrollback wrap heuristic (goes to the owner in the morning). **Red → green (dev-02, production build)** - Red at `e03106d68` (spec only): all 8 new cases failed (focus ×4: "tapping Copy leaves focus in the terminal"; glass ×4: no glass bed), log `r3-red2.log`. - Green at head, all touch + wrap + mouse specs, `--repeat-each=3`: **69 passed (16.3m), 0 failed**, plus the colour unit test 20/20. Log `r3-green.log`. (Two helper bugs in the new spec were fixed on the way: a wait on `document.getAnimations()` that never resolves because some animations are paused, and a strip that sat on the callout's rounded corner; both are in the spec commit, not the product.) - `verify quick`, verbatim: ``` | scope guard | PASS | | migration names | PASS | | Nix Cargo vendoring | PASS | | fmt | PASS | | doc coverage ratchet | PASS | | shape lints off (D351) | PASS | | doc comments on diff | PASS | | fixed test waits | PASS | | ui literals | PASS | | tracer typecheck | PASS | QUICK pre-check passed. This is not the gate: run scripts/verify fast before reporting done. ``` `verify fast` not run (as instructed). dev-02 target removed; my two test servers were stopped by PID. **Infra note:** dev-02's build-slot queue sat stuck for about 30 minutes earlier (a head ticket that was running, slots 2 and 3 free); it cleared by itself.
before-393x852-ink-dense-callout.png
336 KiB
before-393x852-paper-dense-callout.png
307 KiB
after-393x852-ink-dense-callout.png
331 KiB
after-393x852-paper-dense-callout.png
303 KiB
after-375x667-paper-dense-callout.png
295 KiB
after-393x852-ink-copied.png
145 KiB
before-paper-crop.png
106 KiB
after-paper-crop.png
99 KiB
before-393x852-ink-dense-callout.png before-393x852-paper-dense-callout.png after-393x852-ink-dense-callout.png after-393x852-paper-dense-callout.png after-375x667-paper-dense-callout.png after-393x852-ink-copied.png before-paper-crop.png after-paper-crop.png
kayg commented 2026-11-09 06:39:32 +00:00
Author
Owner
Copy link

J404 interim (pre-merge check of the shared popover change) · merged tree 3adae77dc = 3a3466f76 + origin/dev 966c52e07, merge clean, no conflicts (scratch branch j404-merge-426). Nothing merged or deployed. verify fast is still being re-run (see the end).

Specs, production build, dev-02 (~claude/j404-evidence/merged-specs.log): 245 listed, 203 passed, 10 failed, 32 skipped.

  • Touch, wrap and mouse specs on this branch (terminal-touch-select, terminal-selection-wrap): 23/23 pass, 0 fail.
  • Popover and anchored-menu specs: row-menu 5/5, replacement-alignment 4/4, account-card 17/17, palette-navigation 1/1, host-removal-confirm 37/37 (the confirm-tip specs), tooltip-fit 4/4, toast stack: 2 fail (below).
  • 10 failures: docs-palette ×4, palette-polish touch ×2, sign-out ×2 (palette sign-out, D245 storage), toast-stack ×2. The same 10 fail on unmodified dev 966c52e07 (dev-specs.log, same test names, 10 failed / 43 passed in those four files), so they are not caused by this branch.
  • 32 skipped: appwin 15, appwin-card 13, version-pill 4 (the specs' own skips).

Row menu and account card, before (dev) vs after (merged), 1440×900 and 393×852, ink and paper: the element's rect, position (absolute), background, backdrop filter, shadow, radius and parent are identical in all 8 cases (shots/*.json), and a pixel compare inside each popover's own box gives 0 differing pixels in 8/8 (a few hundred pixels differ elsewhere on the page, outside both popovers). Screenshots attached below (before-* / after-*).

**J404 interim (pre-merge check of the shared popover change)** · merged tree `3adae77dc` = `3a3466f76` + origin/dev `966c52e07`, merge clean, no conflicts (scratch branch `j404-merge-426`). Nothing merged or deployed. `verify fast` is still being re-run (see the end). **Specs, production build, dev-02** (`~claude/j404-evidence/merged-specs.log`): 245 listed, **203 passed, 10 failed, 32 skipped**. - Touch, wrap and mouse specs on this branch (`terminal-touch-select`, `terminal-selection-wrap`): 23/23 pass, 0 fail. - Popover and anchored-menu specs: `row-menu` 5/5, `replacement-alignment` 4/4, `account-card` 17/17, `palette-navigation` 1/1, `host-removal-confirm` 37/37 (the confirm-tip specs), `tooltip-fit` 4/4, toast stack: 2 fail (below). - 10 failures: `docs-palette` ×4, `palette-polish` touch ×2, `sign-out` ×2 (palette sign-out, D245 storage), `toast-stack` ×2. **The same 10 fail on unmodified dev `966c52e07`** (`dev-specs.log`, same test names, 10 failed / 43 passed in those four files), so they are not caused by this branch. - 32 skipped: `appwin` 15, `appwin-card` 13, `version-pill` 4 (the specs' own skips). **Row menu and account card, before (dev) vs after (merged), 1440×900 and 393×852, ink and paper:** the element's rect, position (`absolute`), background, backdrop filter, shadow, radius and parent are identical in all 8 cases (`shots/*.json`), and a pixel compare inside each popover's own box gives **0 differing pixels in 8/8** (a few hundred pixels differ elsewhere on the page, outside both popovers). Screenshots attached below (`before-*` / `after-*`).
before-rowmenu-1440x900-ink.png
231 KiB
after-rowmenu-1440x900-ink.png
231 KiB
before-rowmenu-393x852-paper.png
84 KiB
after-rowmenu-393x852-paper.png
84 KiB
before-account-1440x900-paper.png
180 KiB
after-account-1440x900-paper.png
180 KiB
before-account-393x852-ink.png
72 KiB
after-account-393x852-ink.png
72 KiB
before-rowmenu-1440x900-ink.png after-rowmenu-1440x900-ink.png before-rowmenu-393x852-paper.png after-rowmenu-393x852-paper.png before-account-1440x900-paper.png after-account-1440x900-paper.png before-account-393x852-ink.png after-account-393x852-ink.png
kayg commented 2026-11-09 08:54:19 +00:00
Author
Owner
Copy link

J404 final: MERGE-READY · merged SHA tested 3adae77dcc6f3f98966e44865d85666b92bff6cf (3a3466f76 + origin/dev 966c52e07, clean merge, no conflict; scratch branch j404-merge-426, pushed so the VMs could fetch it). Nothing merged or deployed. Result of the check: the shared glassBed / position: fixed popover change breaks nothing else.

1. Specs on a production build, dev-02 (~claude/j404-evidence/merged-specs.log; the specs set their own viewports and themes: 1440×900, 393×852, 375×667, ink and paper)

  • 245 listed: 203 passed, 32 skipped, 10 failed.
  • Touch, wrap and mouse specs of this branch: 23/23 pass.
  • Shared popover and anchored menus: row-menu 5/5, replacement-alignment 4/4, account-card 17/17, palette-navigation 1/1, host-removal-confirm 37/37 (confirm tips), tooltip-fit 4/4, canvas, version-pill, font-picker, account-summary, account-captures, clusters-stray-tooltip: no failure.
  • The 10 failures: docs-palette ×4, palette-polish touch ×2 (label "deploy" has room, 57.87 vs ≥ 58), sign-out ×2, toast-stack ×2. The same 10 tests fail on unmodified dev 966c52e07 (dev-specs.log: 10 failed, 43 passed in the same four files). Not caused by this branch; someone should look at them on dev.
  • 32 skipped: appwin 15, appwin-card 13, version-pill 4 (the specs' own skips; I did not check them on dev).

2. Row menu and account card before (dev) vs after (merged): unchanged (8 cases: 1440×900 and 393×852, ink and paper). Same rect, position: absolute, background, backdrop filter, shadow, radius and parent in every case (~claude/j404-evidence/shots/*.json); pixel compare inside each popover's own box: 0 differing pixels, 8/8. Pairs attached on my earlier comment 88439; all 16 images are in ~/.cache/codex-scratch/j404/evidence/ on the control host.

3. scripts/verify fast, compute01, on 3adae77dc: PASS, VERIFY_EXIT=0 (~claude/j404-evidence/verify-fast.log), verbatim:

+---------------------------+--------------+
| Rust gate                 | Result       |
+---------------------------+--------------+
| Nix Cargo vendoring       | PASS         |
| fmt                       | PASS         |
| migration names           | PASS         |
| migration tooling tests   | PASS         |
| VT native fmt             | PASS         |
| CLI plugin compile-out    | PASS         |
| clippy                    | PASS         |
| SSH runtime test helper   | PASS         |
| real-binary fixture paths | PASS         |
| Traefik component manifest | PASS         |
| release component publication | PASS         |
| test                      | PASS         |
| SSH inode reuse fixture   | PASS         |
| shim process-per-user     | PASS         |
| cargo-deny                | PASS         |
| gate architecture         | PASS         |
| doc coverage ratchet      | PASS         |
| shape lints off (D351)    | PASS         |
| doc comments on diff      | PASS         |
| fixed test waits          | PASS         |
| privileged integration    | SKIP         |
| probe release exclusion   | PASS         |
| nightly fuzz reporter     | PASS         |
| static builds             | PASS         |
+---------------------------+--------------+
VERIFY_EXIT=0

It took three runs, and the first two are not a result:

  • Run 1: the process disappeared mid-run (no VERIFY_EXIT; log verify-fast-run1-died.log). Before it vanished, 4 auth::tests::socket_tests::review_* tests had failed while compute01's load average was 46.
  • Run 2: test and shim process-per-user failed (session::user_host::tests::bulk_turn_advances_sixteen_kib_and_yields_before_the_rest: assertion failed: session.drained; idle_daemon_discovers_unattached_and_adopted_shim_exit_without_requests: an idle-wakeup canary), load average 35 (verify-fast-run2-flaky.log). The branch has no Rust diff at all. Those two tests, run alone, pass 3/3 on the merged tree and 3/3 on unmodified dev 966c52e07 (rust-summary.log). So they are timing tests that fail on a loaded host.
  • Run 3: green, above.
  • Privileged integration is SKIP on this host, as in every gate here.

Not covered: a real iPhone (the system magnifier, Look Up, keyboard behaviour on Copy and double tap); the Copy callout's own visuals (J361 and the two reviewers covered them). I did not check the docs-palette, palette-polish, sign-out and toast-stack failures beyond showing they are the same on dev.

Cleanup: no processes of mine left running; my targets and worktrees on both VMs are removed.

MERGE-READY. Merged SHA 3adae77dcc6f3f98966e44865d85666b92bff6cf. Gates: verify fast PASS (third run); specs 203 passed with 10 failures that are identical on unmodified dev; popovers pixel-identical.

**J404 final: MERGE-READY** · merged SHA tested `3adae77dcc6f3f98966e44865d85666b92bff6cf` (`3a3466f76` + origin/dev `966c52e07`, clean merge, no conflict; scratch branch `j404-merge-426`, pushed so the VMs could fetch it). Nothing merged or deployed. Result of the check: the shared `glassBed` / `position: fixed` popover change breaks nothing else. **1. Specs on a production build, dev-02** (`~claude/j404-evidence/merged-specs.log`; the specs set their own viewports and themes: 1440×900, 393×852, 375×667, ink and paper) - 245 listed: 203 passed, 32 skipped, 10 failed. - Touch, wrap and mouse specs of this branch: **23/23 pass**. - Shared popover and anchored menus: `row-menu` 5/5, `replacement-alignment` 4/4, `account-card` 17/17, `palette-navigation` 1/1, `host-removal-confirm` 37/37 (confirm tips), `tooltip-fit` 4/4, `canvas`, `version-pill`, `font-picker`, `account-summary`, `account-captures`, `clusters-stray-tooltip`: no failure. - The 10 failures: `docs-palette` ×4, `palette-polish` touch ×2 (`label "deploy" has room`, 57.87 vs ≥ 58), `sign-out` ×2, `toast-stack` ×2. **The same 10 tests fail on unmodified dev `966c52e07`** (`dev-specs.log`: 10 failed, 43 passed in the same four files). Not caused by this branch; someone should look at them on dev. - 32 skipped: `appwin` 15, `appwin-card` 13, `version-pill` 4 (the specs' own skips; I did not check them on dev). **2. Row menu and account card before (dev) vs after (merged)**: unchanged (8 cases: 1440×900 and 393×852, ink and paper). Same rect, `position: absolute`, background, backdrop filter, shadow, radius and parent in every case (`~claude/j404-evidence/shots/*.json`); pixel compare inside each popover's own box: 0 differing pixels, 8/8. Pairs attached on my earlier comment 88439; all 16 images are in `~/.cache/codex-scratch/j404/evidence/` on the control host. **3. `scripts/verify fast`, compute01, on `3adae77dc`: PASS, `VERIFY_EXIT=0`** (`~claude/j404-evidence/verify-fast.log`), verbatim: ``` +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | Nix Cargo vendoring | PASS | | fmt | PASS | | migration names | PASS | | migration tooling tests | PASS | | VT native fmt | PASS | | CLI plugin compile-out | PASS | | clippy | PASS | | SSH runtime test helper | PASS | | real-binary fixture paths | PASS | | Traefik component manifest | PASS | | release component publication | PASS | | test | PASS | | SSH inode reuse fixture | PASS | | shim process-per-user | PASS | | cargo-deny | PASS | | gate architecture | PASS | | doc coverage ratchet | PASS | | shape lints off (D351) | PASS | | doc comments on diff | PASS | | fixed test waits | PASS | | privileged integration | SKIP | | probe release exclusion | PASS | | nightly fuzz reporter | PASS | | static builds | PASS | +---------------------------+--------------+ VERIFY_EXIT=0 ``` It took three runs, and the first two are not a result: - Run 1: the process disappeared mid-run (no `VERIFY_EXIT`; log `verify-fast-run1-died.log`). Before it vanished, 4 `auth::tests::socket_tests::review_*` tests had failed while compute01's load average was 46. - Run 2: `test` and `shim process-per-user` failed (`session::user_host::tests::bulk_turn_advances_sixteen_kib_and_yields_before_the_rest`: `assertion failed: session.drained`; `idle_daemon_discovers_unattached_and_adopted_shim_exit_without_requests`: an idle-wakeup canary), load average 35 (`verify-fast-run2-flaky.log`). The branch has no Rust diff at all. Those two tests, run alone, pass 3/3 on the merged tree and 3/3 on unmodified dev `966c52e07` (`rust-summary.log`). So they are timing tests that fail on a loaded host. - Run 3: green, above. - Privileged integration is SKIP on this host, as in every gate here. **Not covered:** a real iPhone (the system magnifier, Look Up, keyboard behaviour on Copy and double tap); the Copy callout's own visuals (J361 and the two reviewers covered them). I did not check the `docs-palette`, `palette-polish`, `sign-out` and `toast-stack` failures beyond showing they are the same on dev. Cleanup: no processes of mine left running; my targets and worktrees on both VMs are removed. **MERGE-READY. Merged SHA `3adae77dcc6f3f98966e44865d85666b92bff6cf`. Gates: `verify fast` PASS (third run); specs 203 passed with 10 failures that are identical on unmodified dev; popovers pixel-identical.**
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
j470b-server-mesh-rejoin
mesh-compat-382
j444b-d381-key-wins-flood
j461b-macos-utun-reader-thread
j475b-remove-host-motion
fix-reconnect-new-link
j481-d380
dev
j479-mesh-soak
fix/j476-host-facts
j475-motion-pill-split
j470-mac-node-join
root-upgrade-d376
j474-review-merge
latency-targets-120hz
nightly-suite-shards
gotatun-pin-j469
macos-utun-reader-thread
fix-flood-echo-stall
wip/j416
motion-fast-ease-out
mesh-addhost-j259
takeover-close-terminals-d368
kbd-instant-followups
fix-trust-key-double-click
nightly-suite-partial-report
appwin-icons-j207
rusqlite-slice1
swipe-content-first-d374
mesh-mtu-381
addhost-instant-open
glass-contrast-d365-mock
glass-contrast-d365
signout-followups-442
fix-startup-toast-red
research-gotatun-matrix-j317
themes-sync-j304
fix-load-flakes
fix-nix-zig-patch
fix-compute01-browser-baseline
ratchet-size-baseline-419
predict-auto-d130
revoke-coverage-d367
nightly-suite-queue-timeout
nightly-browser-suite
fix-rustsec-2026-0459
fonts-ioskeley-obel
log-retention-d375
fix-kbd-no-anim-427
fix-dev-red-specs-2
takeover-proof-305-j205
wip/j410-server-pump
ui-consistency-d372
fix-osc52-clipboard
libghostty-vt-1.0
forget-retired-host-d369
j404-merge-426
fix-setup-lock-race
fix-touch-select-426
revoke-access-report-ui
fix-predictive-echo-specs
fix-selinux-agent-exec
wip/j395-checkpoint2
wip/j395
j391-review-scratch
ui-consistency-j195-deferred-rebased
fix-rustc-paths-421
decisions-wording-341
fix-install-suite-red
fix-dev-rtt-specs
fix-pending-removal-align
fix-privileged-skip
fix-install-suite-404
review-j368
mesh-dns-j254
revoke-access-report-d356b
revoke-privilege-round5
fix-removal-confirm
review-j360
gotatun-rotation-367
deps-refresh-340
review-j354
review-J351
review-j350
addhost-takeover-d359
revoke-report-items-347
revoke-end-sessions-d362
review-j348
size-mesh-ssh-346
size-trim-337
revoke-access-diff-d356
j259-pre-rebase-b06c5a6
j305-ended-at-start-spec
addhost-sudo-consent
fix-layout-real-specs-411
fix-shell-open-411
fix-console-hosts-373
fix-shell-exit
fuzz-host-name-353
fix-host-details-410
macos-filter-j197
research-gotatun-tricks
revoke-p3-j216l
stale-state-pill
revoke-followups-j216k
revoke-progress-j216e
backup/macos01-20261106/lume/j197
backup/macos01-20261106/lume/appwin-icons-j207-stash0
backup/macos01-20261106/claude/stash-0
backup/macos01-20261106/claude/stash-1
backup/macos01-20261106/claude/J80-macos-s1
backup/macos01-20261106/claude/J80
backup/macos01-20261106/claude/J80-slice1-20261102
backup/macos01-20261106/claude/J16b-appwin-on-dev
backup/macos01-20261106/claude/appwin-launch-idle-146
backup/macos01-20261106/claude/J80-s1-run
wip/j259
trust-285-p3c
runtime-lock-j216f
mesh-ssh-p3-j175
tun-proof-p3b
trust-285-p3
terminal-polish-j271
tun-proof-p3
ssh-install-flakes-j216i
fuzz-fixes-j264
trust-285-a
trust-285-b
restart-oversized-fixture-j198
backup/macos01-20261106/lume/appwin-icons-j207
restart-bound-j216h
up-preflight-p3-j198
backup/macos01-20261106/lume/j251
up-preflight-j198
wip/up-preflight-j198-j245-p2
wip/up-preflight-j198-j245-p1
agent-fsync-offloop-j196
wip/up-preflight-j198-j245
settings-offscreen-j231
wip/up-preflight-j198-j226-inspection
wip/j103f
themes-j206
wip/up-preflight-j198-j226
flakes-j216c
backup/macos01-20261106/claude/j197
flakes-j216b
backup/macos01-20261106/claude/j197-red
wip/up-preflight-j198-review
flakes-j216
backup/macos01-20261106/claude/appwin-icons-j207
engineering-j208b
appwin-mock-j209
toast-stack-j165
engineering-j208
decisions-d351
decisions-d350
decisions-d349
appearance-presets-mock-j129
decisions-d348
docs-audit-spec
doc-coverage-baseline-raise
decisions-d347
decisions-d346b
decisions-d346
j88h-profile
ui-consistency-j195-deferred
ui-consistency-j195
engineering-practices-j194
preferred-login-ui-j166
signin-pools-263-j174
macos-node-265-s1
decisions-d345
decisions-d344
proxy-cgroup-j164
verify-quick-fuzz-329
proxy-start-io-j168
timing-flakes-j163
up-263-s1
mcp-events-j147
verify-quick-fuzz-j158
proxy-p3-j159
agent-resource-j157-detect
wip/j103d-final
remote-cancel-hang-j155
proxy-traefik-281-j62d
j149-verify-quick
proxy-traefik-281-j62b
j150-drop-unnamed
wip/j79d-red-r4
wip/j103d-d332
rootless-toast-j142
wip/j103d-head
wip/j103d
j125c-mcp
macos-build-j136
backup/macos01-20261106/claude/j136
wip/j79d-red-cli
wip/j79d-red-server
j121-link-liveness
j118c-tracer-failures
wip/j79c-head
release-desktop-j131
j124c-signed-macos
wip/j79c-red-p3
wip/j79c-red-p23
wip/j79c-red-p22
merge-j128-check
perf-deferred-j128
sidebar-toggle-260
j127-shortcut-changed-dot
j125-mcp
j118-tracer-failures
j126-account-summary
settings-sections-mock-j92
j120-pinch-lock
sidebar-292
backup/macos01-20261106/claude/j120
taste-2026-11-03
row-actions-242-215
backup/macos01-20261106/claude/stash-0-untracked
backup/macos01-20261106/claude/J80-265-s1-run
backup/macos01-20261106/claude/J80-macos-s1-20261102
j112-canvas-commands
ui-no-overshoot-scroll
fix-recovery-restart
backup/macos01-20261106/claude/stash-1-untracked
taste-2026-11-02
size-baseline-ratchet
machine-cards-structure-j108
j109-boot-copy-mockups
fix-dev-red
wip/up-263-s1
j95b-settings-rows
proxy-traefik-281
j95-option-rows
docs-human-j91
docs-keys-palette
fix-flood-1013-270-j38d
mesh-ssh-s1
mesh-ssh-s0
addhost-d291-tunnel
test-stability-load
mesh-relay-220
repro-296
shim-thp-off-24
mesh-pins-258-v2
mesh-dns
signin-d311-merged
docs-pages
ui-size-283
fix-ui-batch-specs
wip/j12b-edge-update
gate-on-arm
fix-rename-sync-293
trust-285-j60b
fix-focus-ring-clip
integ-ui-1101
edge-update-v2
signin-d311
taste-2026-11-01
layout-per-tab-250
ui-approved-splitopt-224
ui-approved-splitopt-224-on-dev
icon-text-centre-282
ui-approved-189-212-216
research-reverse-proxy
transfer-budget-320
review-mesh-pins-258-r2
sidebar-pill-259
review-mesh-relay-220
wip/j12-edge-update-v2
nix-package
status-crash-loop-237
fix-dev-test-failures
fix-fixture-socket-path
wip/mesh-relay-220-pre-dev-advance-rebase
fix-canvas-side-clip
research-head-to-head
frosted-real-256
review-mesh-pins-258
wip/mesh-relay-220-pre-final-rebase
review-layout-per-tab-250-r2
broad-suite-triage
nix-devshell
review-status-crash-loop-237-r3
wip/nix-devshell-pre-p3
taste-frosted-real
wip/code-serve-demo-2026-10-30
wip/code-serve-glass-2026-10-30
wip/code-serve-clusters-2026-10-30
wip/code-merge-release94-dev-2026-10-30
wip/rproxy-delta-scratch-2026-10-30
wip/stale-note-257-staged-2026-10-30
wip/d241-size-staged-2026-10-30
review-broad-suite-207
wip/mesh-pins-258-pre-split
wip/mesh-relay-220-pre-rebase
mesh-measure-220
plan-211-update
review-edge-autoupdate-r3
perf24-reconnect
review-stale-note-r2
release-prod-live
d241-shell-marker
review-mesh-pins-server-258
review-revocation-r3
review-setup-token-r2
review-sync-journal-r3
wip/review-mesh-pins-258-prune-review-1995fee46
frosted-scroll-256
sync-journal-r3
revocation-per-account-wip
review-tokio-asyncfd-0386
review-sync-p3s
wip/stash-8by64-stash0-2026-10-30
mesh-drag-r2
setup-token-data-wip
release-9-4
auth-busy-1013
mesh-view-overlap
sync-p3s
review-auth-busy-wake
review-readd-r8-wip
mesh-drag-bounds
audit-auth-93
app-windows-stream-start
review-authority-cap-r3
app-windows-r7-measure
readd-minimal-r6
readd-minimal-r5
app-windows-r7-soak
readd-minimal-r4
app-windows-r7
app-windows-r6
readd-minimal-r3
readd-minimal-r2
readd-minimal
merge-release8-dev
review-busy-page-liveness-r2
review-busy-page-liveness
mesh-view-trim
review-authority-read-retry-r2
review-remove-uninstall-r2
release-retry-reattach
release-retry
review-remove-uninstall
review-authority-read-retry
app-windows-ship
mesh-autoswitch-p2
mesh-ssh
mesh-relay
mesh-macos-node
screen-and-control
session-labels
review-real-logins-r5
review-real-logins-r4
review-real-logins-r3
shim-memory-r2
review-real-logins-r2
shim-memory
shadow-addhost-instant
shadow-real-logins
shadow-same-session-flood
shadow-prod-perf-fix
hosts-from-server
app-windows-real
app-windows-transport
release-prod-6w
mesh-policy-verify
quiet-close-toast
dev-merge-r4
review-mesh-agent-transport
review-app-windows-transport
settings-roundtrip
review-release-pipeline-r4
mesh-agent-transport
ended-strip-fix
review-audit-ca-update-d244
settings-rows
phone-width
signout-clear
fix-restart-tests
dev-merge-rp
signer-hardening
review-audit-ca-update
terminal-fidelity
review-audit-ca-update-wip
mesh-integrate
review-mesh-agent-transport-515d3712
review-mesh-dataplane-feec1988
mesh-dataplane
signout
research-canvas-node
wip/review-renderer-cost
wip/review-terminal-fidelity
padlock-glyph
review-release-pipeline-2e95b689
phone-polish
sidebar-polish
inbox-generic
review-mesh-dataplane-378d4055
prod-batch
ui-inbox
ui-ca-unlock
ui-machine-sections
review-mesh-dataplane-a5f744fb
tls-session-prod
session-bounds-prod
fix-dead-session
session-bounds
tls-session
palette-polish
shortcuts-d222
research-appmeta
phone-gutter
ui-centre-button
wip/ui-centre-button
wip/tls-session
review-addhost-breakit
ui-platform
renderer-cost
remote-agents
mesh-api
ssh-bootstrap
prod-hotfix2
prod-fixes
review-remote-agents
review-mesh-dataplane
session-life
research-browser
review-mesh-api
research-capture
review-mesh-api-round3
review-mesh-api-round1
gate-stability
review-gate-stability
review-relay-direct
review-mesh-api-findings
proto-sessions
font-picker
proto-addhost
proto-windows
mesh-core
latency-gate
research-bsd
research-discovery
keytray-mods
mock-core
ui-tweaks
fix-motion-sel
fix-motion
integrate2
fix-robust
t-web
t-server
vt-ghostty
fix-snap-echo
t-web-auth
review-vt2
t-agent
lane-protocol
lane-shim
lane-auth
s1-base
s1-contracts
s1-workspace
No results found.
Labels
Clear labels   
launch-qa
Findings from launch QA.

No labels
launch-qa
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
kayg/infraternal#426
No description provided.