Skip to content

omamobile-shell 0.2.119: a swipe up closes the shade card and the Wi-Fi and Bluetooth panels - #553

Open
den-kloot wants to merge 4 commits into
derekross:mainfrom
den-kloot:omamobile-shell-quicksettings-swipe-up
Open

den-kloot wants to merge 4 commits into
derekross:mainfrom
den-kloot:omamobile-shell-quicksettings-swipe-up

Conversation

@den-kloot

@den-kloot den-kloot commented Nov 9, 2026 •

Copy link
Copy Markdown
Contributor

What and why

Quick settings' header says "Tap outside it, swipe it up, or press Escape to put it away", but only the sheet took that swipe:

  • The notification card under the sheet had a bare MouseArea { anchors.fill: parent } that swallowed every touch, so a swipe up that started on the card (most of the screen once a few notifications are there) did nothing at all. The card's MouseArea now moves the sheet's drag.offset, which the card already hangs under (shadeTop follows sheet.y), so both cards follow the finger and close past the sheet's own threshold (Style.space(60)). The notification list only takes vertical drags when it is longer than the card (interactive: contentHeight > height + 1, as sheetFlick already does); before, a short list's Flickable took the swipe and could not scroll anyway.
  • The Wi-Fi and Bluetooth panels are full screen. On the home screen they cover the dock, which owns the swipe up there, so only the back arrow got out of them. A strip at their bottom edge (a DragHandler, y only), as tall as the launcher's home strip (Math.max(44, Style.space(44))), now closes them on a swipe up past 24 px, the same threshold the home strip uses. Over an app the home strip reserves that band, so while a page is up a companion window takes it at the true bottom edge (see the comments).

Nothing else in the sheet changes: tiles, sliders, the pager's sideways swipe and a row's sideways swipe keep their own handlers.

Device and build

  • Phone: Pixel 9a (tegu)
  • Kernel release (uname -r): 7.3.0-rc6-omamobile-push1-gcbb1a7eb0a78 (a local debug build, see below)
  • Packages changed, with versions: omamobile-shell 0.2.119 (0.2.108 when opened; rebased onto main's 0.2.118)

Checks (no phone needed)

  • tools/pr-check upstream/main passes
  • tests/all: two failures, both in tests/kernel-pr-check ("the kernel config doesn't configure", "a new board with a sibling's message"); they need a kernel tree the phone checkout doesn't have, and this change touches no kernel file. Not re-run on main to compare.
  • packages/omamobile-shell/test/all passes (316 ok on the rebased head)
  • omamobile-shell's version bumped to 0.2.119, in the commit subject (rebased onto main at 0.2.118)
  • Commit has an area prefix and a body that says why; AI help is a Co-Authored-By: trailer, no Signed-off-by:

New static checks in test/all: the card follows a swipe up through drag.offset, a short list leaves the swipe to the card, and both panels close on a swipe up from their bottom edge. The gestures themselves are driven on the phone by tools/phone-gesture-test (#652); its results are in the comments.

On the phone

  • I ran this on my own phone: the changed files as local plugin copies (~/.config/omarchy/plugins/omamobile.quicksettings) over omamobile-shell 0.2.100, not a built 0.2.108 package. The shell restarted with no QML errors; quick settings opened and closed from the bar. Since then, the built 0.2.119 package installed as a fresh install, with scripted touches (tools: phone-gesture-test, on-phone gesture tests of the quick-settings pages #652), on the 9a and in an aarch64 VM: see the comments.
  • How it was tested: the agent (Claude Code, on the phone) wrote the change, restarted the shell, checked the journal for QML errors and took screenshots of quick settings open. It cannot swipe, so the gestures are mine (the owner) to try by hand: the problem report was mine ("dragging up or down on the tiles and the notifications moves nothing"), the panels' swipe has been on the phone since this morning, the card's swipe since this afternoon. I will confirm each by hand in a comment.
  • What I did not test: a long notification list (scrolling first, then the swipe from the list's end), landscape (the shade beside the sheet), another phone.
  • Nothing in this change writes outside slot B and userdata, weakens the storage guard, or reads registers from userspace (AGENTS.md §1)

The phone's kernel is a local build (gcbb1a7eb0a78: the 9a daily tree plus an ACPM timeout debug print and the CP-link experiment g7423da0736b1); this change doesn't touch anything they affect.

🤖 Generated with Claude Code

https://claude.ai/code/session_01176vSdGN6JWgiwDLYXe1ES

@derekross

Copy link
Copy Markdown
Owner

Thanks for this. The card change is exactly right: driving the sheet's own drag.offset from the card keeps both cards together and reuses the sheet's threshold, and making the short list non-interactive is the same rule the sheet already has. I also appreciate the clear account of who tested what.

One change before I merge, in the Wi-Fi and Bluetooth pages' new bottom strip. It's 36 px tall at z: 100, and on the phone (no outer gaps, 18 px panel padding) roughly the bottom 18 px of the network list lies under it. I built a small qmltestrunner probe with the same DragHandler over a ListView: a swipe up that starts on the part of the list under the strip closes the page (contentY stays 0), and the same swipe 80 px higher scrolls the list. So a thumb scroll from the last visible network closes the page. Could you keep the strip below the panel's content, for example only the panel's bottom inset (panel.contentBottomInset, plus Style.gapsOut), so the very edge still closes and the list keeps its scroll? The same applies when the keyboard is up, where the strip sits just above it over the connect card.

Two notes, nothing to change here:

  • Both the sheet's swipe and now the card's follow the finger at half speed, because the offset is measured in an item that moves with it (the same probe gives -10, -10, -20, -20, … for a finger moving 10 px at a time), so the 60 px threshold takes about 120 px of travel. That's been in the sheet all along; I'll look at it separately.
  • If omamobile-shell 0.2.107: a Hotspot page behind the tile's chevron #539's Hotspot page lands too, it should get the same edge swipe, so the three pages behave alike.

On versions: #557 also takes 0.2.108, and git merges two identical version lines without a conflict, so I'll merge this one as 0.2.108 and renumber #557 to 0.2.109 myself. After your fix and your hand check, I'll try it on my 7 Pro and merge.

@derekross

Copy link
Copy Markdown
Owner

A correction to my version note above: #542 merged first, as omamobile-shell 0.2.108, so this one will take 0.2.109 when it merges. I'll set that myself, no change needed from you.

@derekross

Copy link
Copy Markdown
Owner

A version note: five shell PRs merged today and main is at 0.2.113 now, so this one will take the next free number when it's ready, not 0.2.109 as I said. I'll set it at merge.

den-kloot added a commit to den-kloot/omarchy-mobile that referenced this pull request Nov 10, 2026
The owner asked that the full-screen pages behind quick settings' tiles
close the same way. The Wi-Fi and Bluetooth pages get a swipe-up strip
at their bottom edge in derekross#553; the Hotspot page now has the same one:
max(28, Style.space(36)) high, a swipe up past 24 px closes it, and the
page's body ends above the strip, so nothing on the page lies under it.

test/all on a PC: 304 ok.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@den-kloot

Copy link
Copy Markdown
Contributor Author

A correction to my own follow-up (fa37750). Cutting the strip down to the panel's bottom inset plus Style.gapsOut kept it off the end of the list, but on the phone the outer gap is 0 (hypr/phone.lua), so the strip was too thin to hit. I found that out on 2026-11-10: a swipe up on the Wi-Fi page did nothing.

f054e06 does it the other way round. The strip is back to Math.max(28, Style.space(36)), and the page's body now ends above it (bottomMargin: Math.max(panel.contentBottomInset, swipeStrip.height)). The end of the list is never under the strip, so a scroll there stays a scroll.

  • test/all on a PC: 301 ok.
  • On my 9a, as a local plugin copy: no QML errors, and in a screenshot the Wi-Fi list ends above the strip.
  • The 36 px strip itself is the one I used by hand before fa37750.

The Hotspot page in #539 gets the same strip (7b0b7a7), so all three pages close the same way.

@derekross

Copy link
Copy Markdown
Owner

Thanks for the honest correction, and for finding by hand that the thin
strip couldn't be hit. The second version is the better design: I read
both pages at f054e06, and the list, the scrim and the connect card are
all inside the body that now ends above the strip, so a scroll from the
last network stays a scroll, and the card with the keyboard up is clear of
it as well. That answers my point from the first round.

What's left before I merge:

  1. A rebase onto main, please. The branch conflicts in the version
    file (main is at 0.2.114 now), and GitHub runs no checks on a
    conflicting pull request, so your last two commits haven't been through
    CI. Any number above main's is fine; I'll set the final one at merge.
  2. Your hand check of the final version. You wrote that the strip is
    the one you used before fa37750, and that the list ends above it in a
    screenshot. Could you confirm three things by hand on this head: a
    swipe up from the bottom edge closes the Wi-Fi and the Bluetooth page, a
    scroll that starts on the last visible network scrolls, and a swipe up
    on the notification card puts quick settings away? The card's swipe is
    the one I haven't seen confirmed yet.
  3. My own try on the 7 Pro, as I said I would: the same three, the
    Wi-Fi password card with the keyboard up, and a long notification list.

Nothing for you to do about this one: your first commit ends with a
Claude-Session: link, which I'll leave out of the squashed message.

This one goes in before #539, so the Hotspot page follows the same strip.

den-kloot added a commit to den-kloot/omarchy-mobile that referenced this pull request Nov 10, 2026
…ed screenshots (public, redacted)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@den-kloot

Copy link
Copy Markdown
Contributor Author

Hardware test — swipe-up to close (Pixel 9a)

Tested on a Pixel 9a, omamobile-shell 0.2.108 from this PR, clean install (no personal customizations). Ten scenarios, each with an annotated screenshot (the gesture path) and a screen recording.

Result: 8/10 as expected. On the home screen everything works. The only misses are the two bottom-edge swipes over an app (07, 08): there the page stays open and the shell's Recents opens behind it — the page's own swipe strip still closes it (09, 10).

# Scenario Expected Result
01 Wi‑Fi page (home) — swipe up from the bottom edge the Wi‑Fi page closes ✅ PASS
02 Bluetooth page (home) — swipe up from the bottom edge the Bluetooth page closes ✅ PASS
03 Wi‑Fi page (home) — drag up starting on the last visible network the page stays open (the list scrolls) ✅ PASS
04 Wi‑Fi page (home) — drag up in the page body, above the strip the page stays open (only the strip closes it) ✅ PASS
05 Quick Settings (home) — swipe up on the notification card header quick settings close ✅ PASS
06 Quick Settings (home) — drag up starting in the notification list the list scrolls, quick settings stay open ✅ PASS
07 Wi‑Fi page over an app — swipe up from the bottom edge the Wi‑Fi page closes ❌ FAIL — stayed open; the shell's Recents opened behind it
08 Bluetooth page over an app — swipe up from the bottom edge the Bluetooth page closes ❌ FAIL — stayed open; the shell's Recents opened behind it
09 Wi‑Fi page over an app — swipe up from the page's own strip the Wi‑Fi page closes ✅ PASS
10 Bluetooth page over an app — swipe up from the page's own strip the Bluetooth page closes ✅ PASS
Per-scenario screenshots and recordings

01 — Wi‑Fi page (home) — swipe up from the bottom edge ✅

Expected: the Wi‑Fi page closes.

01-wifi-bottom-edge-home

01-wifi-bottom-edge-home.mp4

02 — Bluetooth page (home) — swipe up from the bottom edge ✅

Expected: the Bluetooth page closes.

02-bluetooth-bottom-edge-home

02-bluetooth-bottom-edge-home.mp4

03 — Wi‑Fi page (home) — drag up starting on the last visible network ✅

Expected: the page stays open (the list scrolls).

03-wifi-drag-from-last-network-home

03-wifi-drag-from-last-network-home.mp4

04 — Wi‑Fi page (home) — drag up in the page body, above the strip ✅

Expected: the page stays open (only the strip closes it).

04-wifi-drag-in-body-above-strip-home

04-wifi-drag-in-body-above-strip-home.mp4

05 — Quick Settings (home) — swipe up on the notification card header ✅

Expected: quick settings close.

05-notification-card-swipe-home

05-notification-card-swipe-home.mp4

06 — Quick Settings (home) — drag up starting in the notification list ✅

Expected: the list scrolls, quick settings stay open.

06-notification-list-scroll-home

06-notification-list-scroll-home.mp4

07 — Wi‑Fi page over an app — swipe up from the bottom edge ❌

Expected: the Wi‑Fi page closes.
Observed: stayed open; the shell's Recents opened behind it.

07-wifi-bottom-edge-over-app

07-wifi-bottom-edge-over-app.mp4

08 — Bluetooth page over an app — swipe up from the bottom edge ❌

Expected: the Bluetooth page closes.
Observed: stayed open; the shell's Recents opened behind it.

08-bluetooth-bottom-edge-over-app

08-bluetooth-bottom-edge-over-app.mp4

09 — Wi‑Fi page over an app — swipe up from the page's own strip ✅

Expected: the Wi‑Fi page closes.

09-wifi-page-strip-over-app

09-wifi-page-strip-over-app.mp4

10 — Bluetooth page over an app — swipe up from the page's own strip ✅

Expected: the Bluetooth page closes.

10-bluetooth-page-strip-over-app

10-bluetooth-page-strip-over-app.mp4

@derekross

Copy link
Copy Markdown
Owner

Thank you for the test table, the screenshots and the recordings. That
covers what I asked for: the strip closes both pages, a drag that starts on
a row doesn't, and the card's swipe puts quick settings away, which was the
one I hadn't seen. Two notes on reading it, neither a problem: the captions
say the swipes were simulated (uinput), so I'll count the thumb test as the
one you did by hand before fa37750 plus mine; and scenario 03 had three
networks, so nothing could scroll there. I'll try a long list on my phone.

07 and 08 aren't failures of this change. Over an app the page ends
above the home pill (your screenshot shows the pill under the page's
border), and the swipe starts on the pill. That edge belongs to the system
gesture, which opens recent apps, on main as well. What's wrong there is
that recent apps opens with the page still up, and that deserves its own
issue: the home gesture should close an open quick-settings page. Not for
this pull request.

What's left:

  1. The branch still conflicts with main, so GitHub has never run the
    checks on your last two commits. Main moved a lot today and is at
    0.2.117. Only the version file conflicts; merge main in or rebase, as
    you prefer, with any number above 0.2.117. I'll set the final one.
  2. What changed on main under these pages (omamobile-shell 0.2.117, omamobile-settings 0.1.1: Back goes back one level, and home from an app's top level #643): Back is phone-wide
    now. The back gesture sends Escape, and on the Wi-Fi, Bluetooth and
    Record pages Escape goes back to quick settings, like the header arrow.
    Your strip doesn't touch that and merges cleanly next to it. The swipe up
    should keep root.close() as you have it: a swipe up from the bottom
    edge is "home", it leaves; Back goes back one level. So after the merge
    the two gestures do two different things on purpose.
  3. One request: the Record page. It's the third full-screen page over
    the dock, and with this and omamobile-shell 0.2.107: a Hotspot page behind the tile's chevron #539 it would be the only one a swipe up
    doesn't close. Could you give RecordPanel.qml the same strip and bottom
    margin, and add it to the loop in test/all? If you'd rather not, say so
    and I'll do it after.
  4. My own try on the 7 Pro once it's rebased: the strips, a long
    network list, the password card with the keyboard up, the card's swipe
    with a long notification list, and Back next to the swipe.

This still goes in before #539.

den-kloot and others added 4 commits November 11, 2026 04:50
…Fi and Bluetooth panels

Quick settings says "swipe it up to put it away", but only the sheet
took that swipe. The notification card under it had a bare MouseArea that
swallowed every touch, so a swipe up that started on the card (most of
the screen once a few notifications are there) did nothing at all. The
card's MouseArea now moves the sheet's drag.offset, which the card hangs
under, so both cards follow the finger and close past the sheet's own
threshold. The notification list only takes vertical drags when it is
longer than the card; a short list's Flickable took the swipe and could
not scroll anyway.

The Wi-Fi and Bluetooth panels are full screen. On the home screen they
cover the dock, which owns the swipe up there, so only the back arrow got
out of them. A 36 px strip at their bottom edge now closes them on a swipe
up.

test/all checks the three wiring points.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01176vSdGN6JWgiwDLYXe1ES
The 36 px strip lay over the end of the Wi-Fi and Bluetooth lists, so a
scroll started there closed the page. It now covers only the panel's
bottom inset and the outer gap.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ith the page ending above it

fa37750 cut the strip to the panel's bottom inset plus the outer gap, so
it no longer lay over the end of the list. On the phone the outer gap is
0 (hypr/phone.lua), which left a strip too thin to hit: a swipe up on the
Wi-Fi page did nothing (the owner, 2026-11-10). The strip is back to
max(28, Style.space(36)), and the page's body now ends above it
(bottomMargin max(the inset, the strip)), so the end of the list is never
under the strip and a scroll there stays a scroll.

test/all on a PC: 301 ok.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ttom edge over an app

Over an app the launcher's home strip reserves the bottom of the screen
(ExclusionMode.Auto, Launcher.qml's touch: Math.max(44, Style.space(44))),
and the pages' windows fit inside that zone (ExclusionMode.Normal), so the
pages' strip ended above it. A swipe up from the very bottom then started
on the home strip and opened Recents over the page, which stayed open
underneath. On the home screen there is no strip and the page reaches the
bottom, which is why this only showed over an app.

Each page now has a companion window while it is up: anchored to the
bottom, ExclusionMode.Ignore so it reaches the true bottom edge, on the Top
layer above the strip's Bottom layer, and as tall as the page's own strip.
It closes the page on the same swipe, and on a tap, like the scrim. The
page's strip takes the home strip's height (it was Math.max(28,
Style.space(36))), so over an app the companion covers all of the zone the
home strip reserves; with the shorter strip an 8 px band at the top of the
zone still opened Recents. On the home screen the companion lies right
over the page's own strip and the body still ends above it, so it never
reaches the end of the list, the point from the first review. It steps
back while a sheet is up: the password, pairing and rename sheets bring
the on-screen keyboard, and the home strip is the keyboard's strip then.

test/all checks the companion window, its height, and that the page's
strip is the launcher's home strip height.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@den-kloot
den-kloot force-pushed the omamobile-shell-quicksettings-swipe-up branch from f054e06 to 7ce4bbf Compare November 11, 2026 04:06
@den-kloot den-kloot changed the title omamobile-shell 0.2.108: a swipe up closes the shade card and the Wi-Fi and Bluetooth panels omamobile-shell 0.2.119: a swipe up closes the shade card and the Wi-Fi and Bluetooth panels Nov 11, 2026
@den-kloot

Copy link
Copy Markdown
Contributor Author

Rebased onto main (c3cfd86, omamobile-shell 0.2.118), so this is 0.2.119 now (the first commit's subject and the title say so) and CI can run again.

One more change, from testing over an app. A swipe up from the very bottom edge didn't close the Wi-Fi or Bluetooth page while an app was open. The launcher's home strip reserves the bottom 44 px (ExclusionMode.Auto) and the page's window fits above that zone (ExclusionMode.Normal), so the swipe landed on the home strip and opened Recents over the page, which stayed open underneath. On the home screen there is no strip, which is why 01/02 passed in my report above. 7ce4bbf gives each page a companion window while it is up and no sheet is: anchored to the bottom, ExclusionMode.Ignore, on the Top layer above the strip's Bottom layer, closing the page like the strip does. The page's strip now takes the home strip's own height (Math.max(44, Style.space(44)), Launcher.qml's touch; it was Math.max(28, Style.space(36))): with the shorter strip the companion left an 8 px band at the top of the zone that still opened Recents. On the home screen the companion lies over the page's own strip and the body still ends above it, so a scroll from the last network stays a scroll. While a page is up, that band is the page's: a tap there closes it, as a tap outside the panel (the scrim) does, and the home strip's double tap and hold work again once the page is closed. test/all checks the companion, its height, and that the strip is the launcher's height.

Two corrections to my report above. Recents opened over the page, not behind it. And 03 and 04 didn't test what they say: that night only a few networks were in range, the list ended far above the strip, and the "pass" only meant the page stayed open. The test now SKIPs them unless the list reaches the strip, and a page that should stay open needs evidence from the pixels that its content moved up with the drag.

Your three points on this head. The touches are scripted (a uinput touchscreen, points from hyprctl), and the verdicts come from shell call omamobile.quicksettings state, from Recents staying shut, and from that pixel evidence. I ran it on a Pixel 9a with this branch's build installed as a fresh install. I also ran it in an aarch64 VM (Lima on Apple silicon: the phone's packages, Hyprland on vkms at the 9a's 1080x2424 and scale 8/3, the same touchscreen, 20 fictional networks through mac80211_hwsim), because at night the phone sees too few networks for 03/04:

scenario 9a, this head (2 runs) 9a, without 7ce4bbf VM, this head (3 runs) VM, without 7ce4bbf
01/02 bottom edge, home screen PASS – PASS PASS
03 drag from the last visible network SKIP (list too short) – PASS (moved 300-321 px) PASS
04 drag in the body, above the strip SKIP (list too short) – PASS (moved 341-360 px) PASS
05 swipe up on the notification card PASS – PASS PASS
06 drag in the notification list PASS (moved 161 px) – PASS (moved 145 px) PASS
07/08 bottom edge, over an app PASS FAIL (Recents) PASS FAIL (Recents)
09/10 the page's own strip, over an app PASS – PASS PASS
11/12 40 px above the edge, over an app PASS FAIL (Recents) PASS FAIL (Recents)
13 40 px above the edge, home screen PASS FAIL (the list scrolls) PASS FAIL (the list scrolls)

"–": not run on the phone. "Moved" is how far the list's content moved up for a 263 px drag (03, 04) or 300 px (06); 06 stops at the end of the list.

  • A swipe up from the bottom edge closes the Wi-Fi and the Bluetooth page: 01/02 on the home screen, 07/08 over an app, 11/12/13 from 40 px above the edge. Without 7ce4bbf, 07, 08, 11, 12 and 13 fail: over an app Recents opens over the page, and on the home screen (13) the drag scrolls the list.
  • A scroll that starts on the last visible network scrolls: 03 SKIPs on the phone at night, because too few networks were in range for the list to reach the strip (it ended 311-439 px above the page's bottom). In the VM, with 20 networks, 03 and 04 pass on this head: the list moved up with the drag and the page stayed open.
  • A swipe up on the notification card puts quick settings away: 05; and a drag in the notification list scrolls it while quick settings stay (06).

One thing I saw only in the VM, on the earlier 0.2.117 builds: very fast swipes (20 ms, two motion events) sometimes didn't close the Bluetooth page. That happened from its own strip as well, and also without the companion. The Wi-Fi page and the home strip took every such swipe, and at the suite's 300 ms none missed. I haven't chased it.

The test is in #652 (tools/phone-gesture-test, with its arithmetic checked in CI) if you want to run it on the 7 Pro alongside your hand check.

@derekross

Copy link
Copy Markdown
Owner

Thanks for the rebase and for chasing the over-an-app case down to the exclusive zone. I read 7ce4bbf against the launcher: the home strip is on the Bottom layer with its reserved zone, your pages fit above it, and the companion on the Top layer with ExclusionMode.Ignore is the right way to take that band while a page is up. Making the page's strip the launcher's touch height, and having test/all read it from Launcher.qml, closes the 8 px gap properly. CI is green on this head, and all four points from my last comment that were yours are met. Thank you also for withdrawing 03 and 04 yourself: a pass that only means "nothing happened" is exactly what I'd have missed.

I checked the companion's tap and swipe together in a small offscreen probe (the same MouseArea and DragHandler on one item, under touch): a tap closes once, a swipe closes once and doesn't also click, a 20 px drag does neither. One thing the probe showed that explains your VM observation: a swipe that arrives as a single move event does nothing, because the handler becomes active on that move and never sees a translation after it. The home strip has the same shape, so it isn't from this change, and a real panel sends several moves even in a 20 ms flick. I'll open an issue for it.

Nothing more from you is needed for this one. What's left is mine: the run on my 7 Pro (both pages on the home screen and over an app, a long network list, the password sheet with the keyboard up, the card's swipe with a long notification list, and Back next to the swipe). I'll use your suite from #652 alongside my thumb.

Two follow-ups, neither for this pull request:

This merges as 0.2.119, before #539 and #658.

den-kloot added a commit to den-kloot/omarchy-mobile that referenced this pull request Nov 12, 2026
…ithout the password on stdin

Review of derekross#652. A run whose SSH session or USB link dropped got SIGHUP
and exited without its cleanup: do-not-disturb stayed on, the tile layout
stayed aside, the terminal (in its own session) stayed open. SIGHUP now
ends a run through the same cleanup as SIGTERM and Ctrl-C, the cleanup
ignores a second signal and has a timeout on every command, and
results.json is written with "interrupted". Each change is recorded in
~/.local/state/phone-gesture-test/run.json before it is made and the file
goes once everything is back, so what a run that couldn't clean up left
(SIGKILL, a power cut) is put back at the next start or by --cleanup:
do-not-disturb, the tile layout, the terminal, the fictional
notifications, the unread mark (only ever moved back), an open page and
the workspace.

GESTURE_TEST_SUDO_PW is read once and removed from the environment, so
the terminal no longer keeps it. The helper is started with sudo -n after
sudo -S -v: with a ticket from a moment ago, sudo -S didn't read the
password and the helper took it for its first swipe.

Also: the lock and a call's screen are checked before every gesture;
Finger.close() never raises (a sudo child can't be signalled by the
user); missing commands are named with their packages at the start; the
CI test no longer leaves tools/__pycache__; the AGENTS row says where it
has run so far instead of "any device".

On a Pixel 9a (omamobile-shell 0.2.119 from derekross#553): a hang-up mid-run (the
pty's master closed, as sshd does) left nothing changed, where f75cba3
left do-not-disturb on, the tiles aside and the terminal open; a SIGKILL
mid-run was put back by --cleanup and by the next run; with a live sudo
ticket f75cba3's 01 ended in ERROR (the helper parsed the password) and
this one passed; the full suite 11/13 twice (03/04 SKIP: few networks in
range at night), as before.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants