Compare commits

..
Author SHA1 Message Date
Hermenix e80757106f docs(plan): stop the desktop reading as headless Linux — approach and open questions 2026-10-03 23:55:44 +00:00
Hermenix 4b5cb5197a docs+examples: log the cursor marker as open, and film each desk on its own
The marker is now written up as the *open* item rather than a research note: it
is not implemented, the video went out without it, and everything else about the
pointer is exact and readable.

proof-video.py is rewritten for desks, one command, no arguments. It films the
session's own output with `record --output` instead of the published socket --
with desks, the socket carries every output stitched together, so a viewer would
watch two thirds wallpaper. The scrolling beats are real pointer axis events,
down twice and back up once, so the movement cannot be mistaken for a page that
merely loaded a different part of itself.
2026-10-03 19:54:08 +00:00
Hermenix cbf58f6443 docs: two TODO items closed, and the cursor question answered
Closed: containers-do-not-bind and pointer-needs-a-container-or-a-screen. Both
were the same defect seen from two sides, and both are fixed by assigning
windows by pid -- two sessions now hold two outputs with a real browser on each,
and the pointer aims at a desk that has something in it.

The cursor item is rewritten with the answer rather than the speculation: the
compositor draws no cursor image for a virtual pointer, so no capture setting
can make one appear. The evidence is in the file, including the three wrong
answers that came before it and the boring measurement that settled it.

New item: killing a session left two browsers alive, found by counting windows
rather than sessions.
2026-10-03 19:45:42 +00:00
Hermenix acdfb2801e the default capture command now asks for the cursor
grim's `-c` ("include cursors in the screenshot") exists because the screencopy
protocol takes the decision from the client: Hyprland's capture_output carries
an overlayCursor argument and the frame is only composited with the cursor when
it is set. Without it a frame shows that *something* changed and never where it
was clicked -- and a click on the wrong element looks exactly like a click on
the right one, which is the whole reason the pointer work is verifiable at all.

Found by reading Hyprland's Screencopy.cpp rather than guessing at a compositor
setting: `no_hardware_cursors` was already on, and the cursor was invisible
because nobody had asked for it.
2026-10-03 19:42:13 +00:00
Hermenix 352cfdcb24 claim windows by pid, and find the compositor after a restart
Two things stood between "containers enabled" and an application actually
sitting on its own output, and neither was where I said it was.

ipc: discovery refused whenever two instance directories existed, which every
compositor restart leaves behind (a log file where a live compositor used to
be). After two rebuilds the broker could not reach any compositor at all, so no
container could be created and the pointer had nothing to aim at. It now asks
each candidate for its version and uses the one that answers, preferring the
newest only when two answer, and still refusing when none do.

container_host.claim: windows are assigned by the pid the compositor already
reports, not by the class stamp the per-session rule matches -- foot, GTK, Qt
and the chromium wrapper all ignore that stamp, so the desk was created and
left empty while the window sat in special:quarantine. The claim moves the
window to its desk's workspace, makes that workspace the one the desk is
showing, and reads back: the record says where the window is, never where it
was sent. Called from both places a window can appear -- the read-back at start
and container.reconcile -- and skipped in the steady state, because a move is a
real change on screen and should not be dispatched for its own sake.
2026-10-03 19:39:33 +00:00
Hermenix 336474793f docs: the pointer is unusable while containers are off
Found live while trying to demonstrate the scroll verb that was just deployed:
every cursor verb resolves its target through _pointer_output, which requires a
container output, and no session has one -- because containers are off. So the
verb exists and cannot be used, and the acceptance video's missing cursor was
never a capture artefact: nothing could have moved it.

Recorded with the routes out (aim at the shared output when a session owns no
container; or fix the binding first and leave this alone), the cost of the
quick one, and the misleading hint that blames the compositor for a switched-off
feature.
2026-10-03 19:19:24 +00:00
Hermenix 7e0544ed70 docs: point the known-gaps list at the TODO folder, and un-stale its first item
Item 1 said the broker reads none of the container or pointer configuration.
That was true and is not: the settings are parsed beside Policy and a switched-off
container host reports itself unavailable with the key named. The item now
records what actually matters about it -- that containers are off on the
deployment machine because no window can be put in one -- and the section links
docs/TODO/, where each open problem has its impact written down.

The TODO folder itself landed in the previous commit by accident (a `git add -A
docs` swept it in with the scroll work); this commit only makes it findable.
2026-10-03 19:06:17 +00:00
HermenixandClaude Opus 5 6b395305e0 scroll: one verb, two axes, and the sign carries the direction
A session could move a cursor and press it, but not scroll. The protocol
already had the request -- zwlr_virtual_pointer_v1.axis(time, axis, value) --
it had simply never been implemented, in the driver or the broker.

  hhstream session pointer scroll <id> --axis vertical --amount 3

Positive is down (vertical) or right (horizontal): the protocol's direction,
not a choice of ours, which is why direction travels in the sign and there is
deliberately no up/down vocabulary layered over it. The output is whatever the
last warp aimed at, exactly like a button, so scrolling somewhere else is a
warp first -- and the lease is what keeps the whole gesture atomic. Bounded at
100 wheel clicks per request on both sides, because a decimal-point slip should
be a refusal rather than a session scrolled somewhere unrecoverable.

The resume point is re-stamped, not moved: the cursor did not go anywhere, so
the coordinates are re-affirmed and only the stamp advances. Recorded in the
contract as an amendment to the frozen op set, because adding an op to a frozen
interface should be visible rather than implied.

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
2026-10-03 19:05:39 +00:00
HermenixandClaude Opus 5 d0efba0877 examples: the acceptance proof video, as a script
Driving the session's browser over CDP, then a terminal with cowsay, was a
scratch script twice and got pruned from scratch once -- which produced an
80-second video of an idle browser and no error. It is asked for repeatedly, so it
belongs with the rest of the agent-eyes tooling, takes a session id, and says in
its docstring that the URLs are real (the search uses key= singular, the bug that
once silently cost every Gunpost search its filtering).

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
2026-10-03 12:14:57 +00:00
Hermenix 8f3f000783 read the config the module writes: containers can be turned off
The module rendered containers_enabled, workspace_base, container_mode,
max_containers and the pointer lease keys into the TOML, and policy.py read only
[defaults] and [apps] -- so the settings were inert. The defaults coincided,
which made it look honoured, until a deployment tried to turn containers off and
found it could not.

Settings now come from the same file: parsed beside Policy (read by __main__, so
the manager keeps taking a plain ManagerConfig its tests build directly) and
carried into ContainerHost, which reports itself unavailable when switched off.
That is the whole mechanism -- the manager already treats an unavailable host as
'sessions without containers', which is what turning them off means -- and the
reason names the key, so a disabled machine cannot be mistaken for one whose
compositor is down.

This is what made the rollback on the host possible; the machine had containers
built, tested, and impossible to disable.
2026-10-03 12:10:19 +00:00
Hermenix c1dc60af00 hyprland-conf: follow the pointer, because one now exists
vm-pointer measured what this setting actually does: with follow_mouse=0 Hyprland
delivers pointer events only to the keyboard-focused window, so a click aimed at
0.25,0.75 of a session's container put the cursor at x=299 y=519 -- inside that
window -- and both clients reported nothing. The click was aimed correctly and
delivered nowhere.

The comment here said 'no pointer will ever exist; do not let focus depend on
one', which was true when written and is false now that the broker injects a
virtual pointer. The cursor lease already serialises input, so focus following
the pointer cannot make two sessions fight over the keyboard.
2026-10-03 11:50:52 +00:00
Hermenix 1c886eec3f merge surface/cli-module-checks into main
The surface the plan described, and the first composer of real compositor
evidence: seven checks green one at a time, including three new ones. They proved
`hyprctl output create headless` works (PLAN experiment 3, until now
unverified), that a click lands where it was aimed read from the application
itself, and that an unmatched window is quarantined.

They also found three defects no fake could: Hyprland's IPC directory is mode
0700 so the broker could not reach either socket and degraded silently, a broker
restart deleted the driver's socket out from under it, and input:follow_mouse=0
sends pointer events only to the keyboard-focused window, so an aimed click
reaches a client that reports nothing.
2026-10-03 11:50:37 +00:00
31 changed files with 1596 additions and 32 deletions

No files matched your search

+12 -13
View File
@@ -268,20 +268,19 @@ is below. These are defects in code this branch does not own; they are recorded
here because the checks are written against the contract and therefore trip here because the checks are written against the contract and therefore trip
over them. over them.
1. **The broker reads none of the container or pointer configuration.** The deep versions of these — the issue, more than one idea, and why each one
`nix/module.nix` renders `sessions.containers.{max,mode,workspaceBase}` and matters — live in [`docs/TODO/`](docs/TODO/README.md). Update the file there
`sessions.pointer.{leaseTtl,leaseMax}` into the generated TOML as PLAN §10 when one closes, so this list stays a description of what is still true.
names them, and `hhstreamd/policy.py:policy_from_dict` reads only
`[defaults]` and `[apps]`. `hhstreamd/__main__.py:serve` builds its
`ManagerConfig` from command-line flags that have no counterpart for any of
those numbers either, so the broker runs on its own defaults (band 1000,
4 slots, 1280x720, lease 30 s / 300 s).
Those defaults agree with the module's, so a deployment that leaves the 1. **The container host can be switched off, and the broker now honours it.**
options alone behaves as documented; one that changes them is ignored. *(This item previously read "the broker reads none of the container or
`vm-pointer` prints the module's `leaseTtl` beside the TTL the broker pointer configuration." That is fixed: `policy.py` parses those settings
reports, so the discrepancy is visible in the build log rather than beside `Policy`, read in `__main__`, and a host that is switched off reports
inferred. The fix is in `policy.py` and `__main__.py`, not here. itself unavailable with the reason naming the key — so a disabled machine
cannot be mistaken for one whose compositor is down. The reason it matters is
what follows.)* Containers are consequently **off on the deployment machine**:
no window can be put in one yet, and enabling them quarantines every
allowlisted app. That join is the first item in `docs/TODO/`.
2. **`session start --no-container` is refused, not honoured.** PLAN §9 lists 2. **`session start --no-container` is refused, not honoured.** PLAN §9 lists
the flag; `docs/CONTRACT-sessions.md` §4 leaves `start` unchanged, and the flag; `docs/CONTRACT-sessions.md` §4 leaves `start` unchanged, and
+13
View File
@@ -278,11 +278,24 @@ blocked driver must not stall the broker.
```json ```json
{"op": "warp", "output": "hh-x", "x": 0.42, "y": 0.61} -> {"ok": true} {"op": "warp", "output": "hh-x", "x": 0.42, "y": 0.61} -> {"ok": true}
{"op": "button", "button": "left", "state": "down"} -> {"ok": true} {"op": "button", "button": "left", "state": "down"} -> {"ok": true}
{"op": "axis", "axis": "vertical", "value": 3.0} -> {"ok": true}
{"op": "frame"} -> {"ok": true} {"op": "frame"} -> {"ok": true}
{"op": "pressed"} -> {"ok": true, "buttons": ["left"]} {"op": "pressed"} -> {"ok": true, "buttons": ["left"]}
{"op": "release_all"} -> {"ok": true, "lifted": ["left"]} {"op": "release_all"} -> {"ok": true, "lifted": ["left"]}
``` ```
**Amended after the freeze — `axis`, for scrolling.** Added to the op set above
and to the broker's verbs as `pointer.scroll`. Positive vertical is *down* and
positive horizontal is *right*: the protocol's direction, not a choice of ours,
and the reason there is no `up`/`down` vocabulary layered over it. The output is
whatever the last `warp` aimed at, exactly like `button`, so scrolling somewhere
else is a warp first — which is what keeps the whole gesture atomic under the
lease. The magnitude is bounded at 100 wheel clicks per request, in both the
driver and the broker, because a decimal-point slip should be a refusal rather
than a session scrolled somewhere unrecoverable. Unit tests cover validation and
the lease; a real-compositor case is still missing — see
`docs/TODO/scroll-has-no-compositor-proof.md`.
`{"ok": false, "error": "<text>"}` for anything rejected. The driver holds **no `{"ok": false, "error": "<text>"}` for anything rejected. The driver holds **no
lease state** — the broker owns the lease; the driver is an executor that must lease state** — the broker owns the lease; the driver is an executor that must
never be trusted to serialise anything. never be trusted to serialise anything.
+34
View File
@@ -0,0 +1,34 @@
# TODO
Known, unfixed problems — each one real, none of them guessed at.
Each file answers three questions and stops:
- **The issue** — what is actually wrong, with the evidence for it.
- **Ideas** — more than one, and which one is the recommended route.
- **Why it matters** — the impact, not the principle.
| item | severity | state |
| --- | --- | --- |
| [cursor-visibility-in-captures](cursor-visibility-in-captures.md) | high | answered: the compositor draws no cursor for a virtual pointer. Needs a decision on an overlay marker |
| [scroll-has-no-compositor-proof](scroll-has-no-compositor-proof.md) | medium | works live; unproven in a gate |
| [reconcile-while-containers-are-off](reconcile-while-containers-are-off.md) | medium | partly fixed: discovery now survives a restart; teardown still lives inside the feature |
| [kill-can-leave-the-app-alive](kill-can-leave-the-app-alive.md) | medium | found by counting windows, not sessions |
Closed since this folder was written:
- *"Containers do not bind"* — fixed. Windows are assigned by the pid the
compositor reports, moved onto their desk's workspace and read back. Two
sessions now hold two outputs with a real browser on each.
- *"The pointer is unusable while containers are off"* — fixed with it. The
pointer aims at the session's desk, and desks hold applications.
- *"Discovery refuses when two compositor instances exist"* — fixed. Each
candidate is asked for its version and the one that answers is used.
Severity here means "how much of the feature is actually delivered", not "how
loud the failure is". The highest is highest because a whole class of work is
blocked on it: the pointer is exact and invisible, so nothing can show *where*
anything happened.
When one of these is fixed, delete its file and the row — this folder is a list
of what is still true, not a history of what was thought.
@@ -0,0 +1,63 @@
# The compositor draws no cursor for a virtual pointer
## What is open
**The overlay marker is not implemented.** The acceptance video was recorded
without it, so the frames show the pointer's *effects* — a page that scrolled,
a tab that opened — and never the pointer itself. Implement it by stamping
`hyprctl cursorpos` into each frame, mapped into the stitched layout, and label
it as an overlay wherever the frame is published. Nothing else about the pointer
is blocked: position, clicks and scrolling are all exact and all readable.
## The issue — answered, and it is not a capture setting
The pointer moves exactly where it is told, and **nothing is drawn**. Every
plausible cause was eliminated by test rather than by argument:
| candidate | test | result |
| --- | --- | --- |
| `grim` not asking for the cursor | Hyprland's `Screencopy.cpp` takes an `overlayCursor` argument, so `-c` was added to the capture command | no change at the pointer |
| hardware cursor plane | `cursor:no_hardware_cursors` | already `true` |
| no cursor theme installed | installed `adwaita-icon-theme`, present at `/run/current-system/sw/share/icons/Adwaita` (63 cursor files) | no change |
| theme not reaching the cursor manager | `hyprctl setcursor Adwaita 24` at runtime | no change — and note `env = XCURSOR_THEME` in the compositor's own config sets variables for **launched programs**, not for the compositor, which is where my first attempt put it |
| hyprcursor rather than X11 themes | `cursor:enable_hyprcursor` | already `false`: the X11 path |
The decisive measurement: park the pointer at two points that both miss the
browser's toolbar, capture each, diff them — **zero differing pixels**. The only
differences that ever appear are in the URL bar, which is a blinking text caret.
The cause is the protocol. `zwlr_virtual_pointer_v1` has no request to set a
cursor image, and Hyprland renders a cursor only for a pointer that has an image
to draw. A virtual pointer warps the position and has no surface behind it, so
the compositor has nothing to composite: every capture is faithful, and the
cursor is simply not in it.
## Ideas
1. **Stamp the position into the frames, labelled as an overlay.** *(recommended)*
The frame server can ask `hyprctl cursorpos` — this machine answers it exactly
— map it into the stitched frame, and draw a marker there. It is not fiction:
it is the compositor's own answer, drawn where the compositor says the pointer
is, precisely because the compositor draws nothing. It has to be *named* as an
overlay in the code and in anything published from it.
2. **Leave the pixels alone and prove motion by readout.** `hyprctl cursorpos` is
exact: 0.8 across a 1280-wide desk came back as 3584 with the desk starting at
2560. Gates can assert position with no image at all — cheaper and stronger
than a pixel check.
3. **Ask upstream, or try a newer compositor.** Whatever makes a cursor appear
for a virtual pointer is the real fix. Not attempted.
4. **A synthetic pointer client.** Impossible for a plain client (it cannot draw
over another window); it would need a layer-shell surface, which is more
machinery than (1) for the same result.
## Why it matters
Everything about the pointer is verifiable except the part a person can see. That
a click landed *here* is provable by readout and by what changed, but a viewer
cannot tell **where** it happened — and a click on the wrong element and a click
on the right one both produce a screen that changed.
Worth recording: this took three wrong answers (`-c`, the theme, the compositor
setting) before the right one, and two of those came from reasoning rather than
measurement. What settled it was boring — capture the same screen twice with the
pointer in two places and diff them. That should have been the first move.
+42
View File
@@ -0,0 +1,42 @@
# Killing a session does not always kill the browser it started
## The issue
Three chromium processes outlived the sessions that had launched them, while the
broker reported those sessions as ended. Found by counting windows on the
compositor after a probe run: two windows left on the main screen whose sessions
no longer existed, with pids older than the session that had replaced them.
The kill path does signal the process group — the audit line says
`sent SIGTERM to process group 2966561` and that session's process *did* die —
so this is not "the kill does nothing". It is narrower: two of three killings
left a live browser behind, and the two that leaked were started while the
compositor was unreachable from the broker, so they had no container recorded at
all. That correlation is the only clue so far and it is not proof.
## Ideas
1. **Check whether the leak tracks the container record.** *(recommended first)*
Kill a session whose container is `owned` and one whose container is `null`,
several times, and count survivors. If the leak is limited to containerless
sessions, the fix is in the kill path's cleanup of a session that never got a
desk, not in the signal path.
2. **Re-read the survivors from the compositor after every kill.** The teardown
already has `container_survivors` for windows; the same shape applies to
processes, and `procmon` already knows how to identify them by
`(pid, starttime)`. A kill that leaves anything alive should say so in its
reply rather than returning a clean record.
3. **Make the idle reaper the backstop it already is.** A session that was
reaped instead of killed goes through the same teardown; if the leak is in
the exit path, the reaper may be leaking too — worth measuring once (1).
4. **A test that counts processes, not sessions.** Every existing assertion is
about the broker's view. One that counts the app's processes after a kill
would have caught this the first time it happened.
## Why it matters
Processes that outlive their sessions are exactly what the registry's whole
identity design (boot id, start time) exists to prevent, and the ones found were
still holding windows on the compositor — invisible to `session list`, visible
on screen. A machine that accumulates them slowly fills up with browsers nobody
owns, each holding a desk slot if the band is what limits sessions.
@@ -0,0 +1,48 @@
# Cleanup depends on the feature it is cleaning up after
## The issue
Start-up reconciliation for stray `hh-*` outputs lives inside the container
host, so it runs **only when containers are enabled**. Turning containers off
therefore also turned off the broker's ability to see what it had already left
on the compositor.
That is not theoretical: an output created while the feature was on survived the
rollback switch. With containers off, nothing reconciled it, and the frame
capture stayed **2560×720 instead of 1280×720** — the extra width being the
orphan sitting beside the real one. The symptom is a *capture* problem, which is
where it would be looked for, and it was fixed by restarting the compositor,
which drops every session.
## Ideas
1. **Reconcile always, not conditionally.** *(recommended)* On start-up, list the
compositor's outputs, find `hh-*` ones that no live session owns, and remove
them — regardless of whether containers are enabled. "Disabled" should mean
*create none*, not *leave the machine as it was found*.
2. **Never start a session onto an unowned output.** If a leftover is found that
cannot be removed, refuse to place a session on it rather than silently
sharing it — the ambiguity is what makes the capture width unexplainable.
3. **Record removals in the audit trail.** The `hyprctl output remove` calls are
already logged individually; a reconcile pass that removes three outputs
should say so once, with the names, so a later "why is my screen narrower"
has an answer.
4. **Make it a check.** A VM case that creates a container, disables the feature,
and asserts the output is gone — the rollback path deserves the same gate as
the forward path, because it is the one that runs when something is already
wrong.
## Why it matters
A feature that cannot be turned off safely is worse than one that was never
turned on, and this one fails asymmetrically: the *forward* path is tested
(containers are created and torn down in `vm-sessions` and `vm-containers`),
while the *rollback* path was never exercised before it was needed. That is
precisely when it is used.
It also cost a compositor restart — the bluntest tool available, and one that
kills every session — to clean up a state the broker itself had created and
could no longer see. The general rule worth keeping: **teardown code belongs
outside the thing being torn down**, or there will be exactly one configuration
where it does not run, and it will be the configuration you are in when you
need it.
@@ -0,0 +1,50 @@
# Scrolling works, but nothing proves it against a real compositor
## The issue
`hhstream session pointer scroll <id> --amount N` now exists end to end: CLI →
`pointer.scroll` → the lease check → driver `axis` op →
`zwlr_virtual_pointer_v1.axis(time, axis, value)` → `frame`. Validation is
covered by unit tests, and the lease behaviour matches the other gestures.
What is **not** covered is the half that matters: no test scrolls a real
application on a real compositor. The `vm-pointer` gate has no axis case, and
the fake compositor can only prove the wiring — that the broker said the right
thing to the driver, not that a window moved.
Also unproven, and worth knowing before trusting it: whether an app on a
**headless** output honours a bare `axis` event at all. GTK and Chromium
smooth-scrolling both look at `axis_source` and `axis_discrete`, which the
driver does not send today. If they are required, the symptom will be a scroll
that silently does nothing — indistinguishable from a page that would not move.
## Ideas
1. **An axis case in `vm-pointer` that reads the app's own state.** *(recommended)*
Scroll a real app and assert *its* state changed, not the pixels — the
contract already sets this bar for clicks ("verified by reading the app's own
state"). A `foot` running a script that appends a line on scroll, or a page
whose scroll offset the app can report, keeps the assertion honest.
2. **Send `axis_source` and `axis_discrete` alongside `axis`** if (1) shows the
event arriving and the app ignoring it. Small, confined to the driver's
`axis` op.
3. **Keep the direction convention in one place.** Positive vertical is down and
positive horizontal is right, per the protocol. It is documented in
`hhstream_pointer/buttons.py` and mirrored in the CLI's `AXES` — if a third
copy ever appears, that is the point to notice, because two vocabularies over
one sign is how a scroll ends up backwards.
4. **A live check on this machine, not only in a VM.** The video that prompted
this work is a real scroll on a real app; making that reproducible as a
command (like `examples/proof-video.py`) is worth as much as the VM case.
## Why it matters
An unimplemented feature is honest; a **plausible** one is not. A scroll that
does nothing, a scroll that goes the wrong way, and a scroll the app ignores all
present as "the page would not move", and all three would be found by a human
watching a video — which is not a test.
This is the same shape as the bug that made `foot` exit 230 with every test
green: the tests exercised what was built, and the built thing was fine, and the
app exited anyway. Anything whose acceptance proof runs through a real window
needs a real window in the test.
+126
View File
@@ -0,0 +1,126 @@
# Make the Hyprland session look like a normal Linux desktop
Plan only. Nothing here was edited, rebuilt, or measured live. The diagnostic shell (`hyprctl monitors`, `systemctl cat`, `/dev/dri`, portals, audio) was cancelled before it ran. Every live fact below is either from files that were read or from older Chromium logs already on disk. Anything not in those sources is marked unverified.
## What was read
- `/etc/nixos/hyprland-stream.nix` — outputs, capture, session allowlist, fonts, cursor theme, window rule
- `/etc/nixos/common.nix` — QEMU guest, `chromium` on `systemPackages`, `hermes` groups, no audio/portal/timezone block
- `nix/module.nix` — compositor environment, `WLR_*`, `SEATD_VTBOUND`, session broker, what apps inherit
- `nix/hyprland-conf.nix` — generated `hyprland.conf` (cursor, input, `disable_xdg_env_checks`)
- `src/hhstream/capture.py` — `grim` to stdout; capture does not name an output
- `src/hhstreamd/containers.py` — `hyprctl output create headless hh-<session-id>`
- `src/hhstreamd/launcher.py`, `src/hhstreamd/compositor.py` — app environment
- `examples/agent-eyes` — `run` only sets `XDG_RUNTIME_DIR`, `WAYLAND_DISPLAY`, `XDG_SESSION_TYPE`
- `docs/ARCHITECTURE.md` (full). `docs/CONTRACT.md` and `docs/CONTRACT-sessions.md` were searched, not read end to end. Container output name `hh-<session-id>` is frozen in the sessions contract.
- `/var/lib/hermes/.hermes/skills/devops/agent-eyes/SKILL.md` and `headless-wayland-compositor/references/browser-fetch-and-cdp.md`
The SwiftShader flags are **not** in that skill, the allowlist, or `examples/agent-eyes`. They appear in one recorded command (`~/.hermes/logs/process-results/proc_a2b2e2384799.json`): `agent-eyes run -- chromium … --use-gl=angle --use-angle=swiftshader --enable-unsafe-swiftshader … https://discord.com/register`. Older Chromium logs (`proc_53cbb1c64842.json`, `proc_ecc7dddf484a.json`, `proc_3edb9a456ba3.json`) show `virtio_gpu_drv_video.so` init failure, `WebGL1 blocklisted`, and UPower `DisplayDevice` `ServiceUnknown`. Those logs still mention `XDG_RUNTIME_DIR=/run/hyprland-headless-stream` (the old path). The runtime dir in the current module is `/run/hypr-stream`.
## 1. Signal inventory
A page cannot see Hyprland output names, kernel input names, or `XDG_*`. Those matter only when they change something the page *can* read (WebGL, `screen`, media devices, `Intl`).
| Signal | Evidence | What a page actually sees | Live check |
|---|---|---|---|
| SwiftShader forced on | Ad-hoc `agent-eyes run` argv above. Not in the Nix allowlist, so `hhstream session start chromium` would reject those args. `agent-eyes run` execs them as-is. | `WEBGL_debug_renderer_info` / WebGPU adapter info. Stock SwiftShader strings are `Google SwiftShader` and device id `0x0000C0DE`. **The exact string on this box was not read back.** | Unverified |
| WebGL context dies without those flags | Same logs: `ContextResult::kFatalFailure: WebGL1 blocklisted` and a failed GPU command buffer | `getContext('webgl')` is null. That is its own automation signal. | From old logs, not a new probe |
| Virtio GPU present, video init fails | `libva error: …/virtio_gpu_drv_video.so init failed`, `vaInitialize failed`. `hardware.graphics.enable = true`. This is a QEMU guest (`common.nix` imports `qemu-guest.nix`). | Hardware decode missing (`mediaCapabilities`). The GL renderer string without SwiftShader flags was **not** captured. Likely Mesa virgl or llvmpipe, not NVIDIA/AMD/Intel. | Renderer string unverified |
| `/dev/dri` missing | **Not supported.** Architecture says the compositor will not start without a render node, and the VAAPI path found a virtio-gpu driver. | Nothing, if the node exists. | `ls /dev/dri` was not run |
| `hermes` cannot open the render node | `hermes.extraGroups` is `wheel`, `systemd-journal`, and `hhstream`. The compositor user gets `video` and `render`. Chromium runs as `hermes`. | Browser falls back to software GL or a blocklisted WebGL context. This is the likely reason SwiftShader was added by hand. | Groups were not read from the running session |
| Main output name | Module does **not** create the main output. The monitor rule uses an empty name (wildcard). Comments say `Virtual-1` on virtio, `eDP-1` on a laptop. `WLR_BACKENDS=headless` is set and documented as ignored by current Hyprland (aquamarine). | Not visible on `window.screen`. `getScreenDetails().label` might echo the wl_output name, and only after a permission prompt. **Not verified in this Chromium.** | `hyprctl monitors` was not run. Do not assume the name is `HEADLESS-1`. |
| Per-session output | `containers.enable = true` (trial, dated 2026-10-03). `containers.py` runs `hyprctl output create headless hh-<id>`. Contract freezes that name. | Same as above, and only if the window is actually on that output. The Nix comment says Chromium ignores the app-id stamp, so the window may miss the container. | Unverified whether any `hh-*` output exists now |
| Screen 1280×720 @ 30 Hz, scale 1 | `outputs` in `hyprland-stream.nix`; container default `1280x720`. Window rule `size 1040 560`. Launch recipe `--window-size=1040,560`. | `screen.width/height`, `devicePixelRatio` 1. Refresh rate is not on `window.screen`. 720p is a weak signal, not a headless label. | Mode unverified live |
| No real input devices | `WLR_LIBINPUT_NO_DEVICES=1`. Pointer is `zwlr_virtual_pointer_v1` (`hhstream-pointer`). Generated config sets `cursor.invisible = true`. | A page does **not** see kernel or libinput names. Compositor-delivered clicks are real Wayland events; `isTrusted` was not measured. CDP input (`Input.insertText` in the skill) is a separate, stronger tell. No keyboard injector exists (`wtype` / `ydotool` are called out as not installed). | Device list unverified |
| `navigator.webdriver` / headless UA | Allowlist and the skill do **not** pass `--enable-automation` or `--headless`. They do pass `--remote-debugging-port` and `--remote-debugging-address=127.0.0.1`. | `navigator.webdriver` should be false and the UA should not contain `HeadlessChrome`. **Not evaluated in a page.** The open DevTools port is not visible to the origin. Attaching CDP and typing through it is. | Unverified |
| Runtime dir not owned by the app | Compositor `XDG_RUNTIME_DIR=/run/hypr-stream`, owned by `hhstream`. Broker and `agent-eyes run` point the app at that same directory. Old log: `XDG_RUNTIME_DIR (…) is not owned by us (uid 1000), but by uid 994`. | Not visible as a string. Chromium then skips Pulse, portals, and dconf. That shows up as empty media devices and failed permission UIs. | Old path only; current path not re-checked |
| Desktop env not set | App env from `CompositorSession.env()` is only `XDG_RUNTIME_DIR`, `WAYLAND_DISPLAY`, `XDG_SESSION_TYPE=wayland`. Compositor also sets `HYPRLAND_NO_SD_VARS=1`, so it does not publish a desktop session. `HOME` for the compositor unit is `/run/hypr-stream`; for broker apps it is the hermes home. | Pages do not read `XDG_*`. GTK/Qt and Chromium's own desktop integration do. | Unverified in `/proc/<pid>/environ` |
| No portal | `disable_xdg_env_checks = true` because "the XDG portal environment" is not there. Neither Nix file enables `xdg.portal`. | No direct JS API. Clipboard, file choosers, and screencast fail. Not the Discord signup check. | Service state unverified |
| No audio server | No PipeWire, Pulse, or WirePlumber in either Nix file. | `enumerateDevices()` has no `audioinput` / `audiooutput` (kinds can leak before permission). | Unverified |
| No UPower | Chromium log: `org.freedesktop.UPower` `ServiceUnknown`. No upower module in the files read. | `navigator.getBattery()` fails. A tower desktop also has no battery, so this is a weak "VM" tell and a bad thing to fake as a laptop. | From old logs |
| Fonts | `fonts.packages = [ pkgs.dejavu_fonts ]` only, added so `foot` has a monospace. | Canvas / `document.fonts` look like a minimal Linux install, not specifically headless. Emoji and metric-compatible faces are absent. | `fc-list` was not run |
| Locale and timezone | Neither file sets `i18n.defaultLocale` or `time.timeZone`. The session unit does not set `LANG` or `TZ`. | `Intl.DateTimeFormat().resolvedOptions().timeZone` and `navigator.languages`. A VM left on UTC is a datacenter tell if the operator is not in UTC. **The live zone was not read.** | Unverified |
| Fresh profile every launch | Skill and the Discord command use `--user-data-dir=/tmp/eyes-profile-$$` (or a one-off `/tmp` dir). | Empty cookies, storage, and history. Behavioral, not a renderer string. | n/a |
Not listed, because nothing read supports them: a kernel input device whose name contains "headless"; `navigator.webdriver === true`; a UA containing `HeadlessChrome`; a missing `/dev/dri`.
## 2. What we can change
**Stop forcing SwiftShader.** Do not add those three flags to `sessions.apps.chromium.argsAllow`. `agent-eyes run` bypasses the allowlist, so a wrapper has to be the binary both paths exec. Wrapper lives in the NixOS module (`sessions.apps.chromium.exec`) and the skill's launch recipe points at it. Flags to keep: `--ozone-platform=wayland`. Flags to refuse: `--use-angle=swiftshader`, `--enable-unsafe-swiftshader`, `--enable-automation`, `--headless`. Do not spoof `UNMASKED_RENDERER`. Overriding `getParameter` in the page is itself detectable.
**Let Chromium open the same render node the compositor uses.** Add `video` and `render` to `users.users.hermes.extraGroups` in `hyprland-stream.nix`, and restart `hhstream-sessions` so the broker picks up the groups. An already-open SSH session will not see them until a new login. `agent-eyes run` from that SSH session still needs a new login or `sg`. This is the best chance of a real Mesa renderer (virgl) instead of SwiftShader or a null WebGL context.
**Give apps a runtime dir they own.** `CompositorSession.env()` and `examples/agent-eyes` `cmd_run` should set `XDG_RUNTIME_DIR` to hermes' `/run/user/<uid>` (linger is already on) and put a symlink to the compositor's `wayland-N` socket there. Toolkits resolve `WAYLAND_DISPLAY` against `XDG_RUNTIME_DIR`, so the symlink is required. Also set `XDG_CURRENT_DESKTOP=Hyprland` and `XDG_SESSION_DESKTOP=Hyprland` on the app, not by clearing `HYPRLAND_NO_SD_VARS` on the compositor unit. The frame server and `grim` stay on `/run/hypr-stream`.
**Desktop services, after that runtime-dir fix.** In `hyprland-stream.nix` (session-specific, not the whole guest image):
- PipeWire + WirePlumber. No sound card was verified; a null sink is the honest stand-in.
- `xdg.portal` with the GTK portal. The Hyprland portal only after it can see this compositor's IPC, which is not the user session's.
- `upower` so the D-Bus name exists. Do not invent a charge percentage.
- Fonts next to DejaVu: a sans/serif metric set (Liberation or Noto), plus Noto Color Emoji.
**Timezone and locale** belong in `common.nix` (`time.timeZone`, `i18n.defaultLocale`), set to the operator's real zone. The files do not say what that zone is. Do not guess.
**Resolution, only if capture still fits.** 1920×1080@60, scale 1, in `outputs` and `sessions.containers.mode`. Keep the window smaller than the output (the wallpaper is intentional). JPEG q80 at 1280×720 was measured at ~46 ms in the Nix comments; 1080p is about 2.25× the pixels. If `grim` goes much past that, stay at 720p. Do not set `outputs[].name` to `DP-1` or `HDMI-A-1`: a monitor rule whose name matches nothing is silently ignored and the mode never applies.
**Do not rename `hh-*` outputs.** The sessions contract, `grim -o`, pointer binding, and reconcile all use that name. A page almost certainly never reads it.
**One durable Chromium profile** for the signup browser, under `/var/lib/hermes/hhstream-sessions/`, not `/tmp/eyes-profile-$$`. Keep per-run profiles only when two browsers must run at once (Chromium exits if the profile is locked).
**Leave the cursor config alone** in this work. `invisible = true` is in the generated conf and the extraConfig does not turn it off. That affects captures, not fingerprinting.
## 3. What cannot be faked
There is no physical GPU on this guest. The GL string will be Mesa on virtio-gpu, or llvmpipe if virgl does not come up. Both are software-ish VM strings. SwiftShader is worse because `0x0000C0DE` is a known headless-Chrome id. Claiming Intel, AMD, or NVIDIA from stock Chromium is not possible, and a patched Chromium or a JS lie is a louder signal than llvmpipe.
There is no monitor EDID and no physical seat. Renaming an output `DP-1` does not create one. Websites do not read EDID.
`navigator.getBattery()` failing is normal for a desktop tower. A fake laptop battery would disagree with the rest of the machine.
WebRTC can still expose the guest address `10.77.0.2` and the hostname `hermes` (`common.nix`). That is a VM tell, not a "headless Linux" string. Confirm with a page before adding a WebRTC policy flag; do not add it speculatively.
Typing still goes through CDP unless a keyboard injector is added later. Pointer events from `zwlr_virtual_pointer_v1` are the less detectable input. This plan does not add a keyboard.
Discord's phone wall was not reproduced here. The Google signup note (`web-account-signup/references/google-signup.md`) records the same class of wall on this desktop, and it does not say which signal Google used. Fixing the rows above removes the signals we can point at. It does not guarantee the wall goes away.
## 4. Staged plan
**Stage 0 — baseline, no changes.** Small. Run, and save the output:
```bash
sg hhstream -c 'XDG_RUNTIME_DIR=/run/hypr-stream HYPRLAND_INSTANCE_SIGNATURE=$(basename /run/hypr-stream/hypr/*) hyprctl -j monitors'
sg hhstream -c 'XDG_RUNTIME_DIR=/run/hypr-stream HYPRLAND_INSTANCE_SIGNATURE=$(basename /run/hypr-stream/hypr/*) hyprctl -j devices'
ls -l /dev/dri
id hermes
systemctl is-active hyprland-headless hyprland-headless-stream hhstream-sessions hhstream-pointer
```
Then one windowed Chromium **without** SwiftShader flags, and one CDP `Runtime.evaluate` of: WebGL vendor/renderer (or null context), `navigator.webdriver`, `navigator.userAgent`, `screen.width/height/devicePixelRatio`, `Intl.DateTimeFormat().resolvedOptions()`, `navigator.languages`, `enumerateDevices()` kinds, and `getBattery()`'s rejection. That sheet is the only acceptance test that matters.
**Stage 1 — wrapper, stop SwiftShader.** Small. Nix wrapper as `chromium`'s `exec`; skill recipe updated; allowlist does not grow those flags. Proof: `/proc/<pid>/cmdline` has no `swiftshader`; the Stage 0 WebGL expression no longer returns SwiftShader. If the context is null, Stage 1 is incomplete and Stage 2 is the fix, not turning SwiftShader back on.
**Stage 2 — `render`/`video` for `hermes`.** Small config, needs a new login and a broker restart. Proof: same WebGL expression returns a Mesa/virgl (or llvmpipe) string, and the log has no `WebGL1 blocklisted`.
**Stage 3 — user-owned `XDG_RUNTIME_DIR` plus desktop variables.** Medium. `compositor.py` and `agent-eyes` `cmd_run`. Proof: Chromium's stderr does not contain `not owned by us`; a launched `foot` prints `/run/user/<uid>`. Run `checks.vm-sessions` and `checks.vm-pointer` (one at a time; they are timing-sensitive). Failure mode to watch: `Failed to connect to Wayland display`.
**Stage 4 — PipeWire null sink, GTK portal, upower, fonts, real timezone.** Medium, NixOS only, after Stage 3. Proof: `wpctl status` shows a sink; `busctl status org.freedesktop.portal.Desktop` and `org.freedesktop.UPower` are active; `fc-match sans` and `fc-match emoji` hit the new faces; the page's `enumerateDevices()` includes `audiooutput`; Chromium's log no longer says UPower `ServiceUnknown`; `Intl` timezone matches `timedatectl`.
**Stage 5 — 1080p@60, gated.** Small, only if Stage 0's page probe and a `grim` timing say 720p@30 is worth the CPU. Proof: `hyprctl monitors` shows the new mode, and one timed `grim -c -t jpeg -q 80` stays near the old ~46 ms budget. If it does not, revert the mode and keep 720p.
**Stage 6 — durable profile.** Small, convention plus the wrapper's `--user-data-dir`. Proof: second launch reuses the same directory and still starts; cmdline has no `/tmp/eyes-profile`.
Land 0 → 1 → 2 → 3 → 4 → 6. Stage 5 is optional and independent. Do not combine Stage 3 with a resolution change; a dead Wayland socket and a slow `grim` would be one undifferentiated failure.
## 5. Risks
- **Wayland socket.** Stage 3 is the BUG-01 area. The frame server must keep `XDG_RUNTIME_DIR=/run/hypr-stream`. Moving the compositor's own runtime directory re-breaks the 25-character `sun_path` limit on the event socket.
- **`hh-*` names.** Renaming them breaks capture-by-output, the pointer's per-output device, and reconcile. Leave the contract alone.
- **Monitor-rule names.** A non-empty `outputs[].name` that is not the real connector is ignored. The session would come up at the connector's preferred mode and the 1280×720 assumption in capture, the window rule, and the tests would be wrong.
- **Containers trial.** `containers.enable = true` today can park Chromium on the quarantine workspace, so the published frame is an empty desktop. A fingerprint change that assumes the window is on the main output will be "verified" against the wrong screen.
- **CPU.** `debug:vfr = false` renders every output every frame, and each container is another output. 1080p plus two containers is the capture-budget failure.
- **Groups.** `extraGroups` does nothing for a broker that is already running, and nothing for an SSH session started before the rebuild.
- **Allowlist vs `agent-eyes run`.** Fixing only `argsAllow` does not stop the helper from passing SwiftShader. The wrapper has to be on both paths.
- **Cursor and `grim -c`.** Do not flip `cursor.invisible` in the same change. Capture tests do not own that behavior.
- **VM checks.** Run `vm-sessions`, `vm-pointer`, and `vm-hyprland` one at a time after Stages 3 and 5. Two guests at once on this machine flake.
The defensible outcome is a windowed Chromium on virtio-gpu Mesa, with a normal runtime dir, a sink, fonts, and the operator's timezone. It will still be a QEMU guest. It will not be SwiftShader, and it will not be named like a headless Chrome.
+214
View File
@@ -0,0 +1,214 @@
#!/usr/bin/env python3
"""The acceptance video: an SKS search, real scrolling, tabs, and cowsay.
One command, no arguments, and it cleans up after itself. It starts a browser
session (which gets a desk of its own — its own output and workspace), films
*that desk alone*, drives the search over the browser's own HTTP endpoints, and
scrolls the results page with the real pointer -- `hhstream session pointer
scroll`, which sends wl_pointer axis events rather than pressing a key.
Then it kills the browser, starts a terminal session on its own desk, and films
cowsay saying what it says. The two films are joined into one mp4.
Why `--output` and not the published frame socket: with desks, the socket
carries every output stitched side by side, so a viewer would watch two thirds
wallpaper. `--output` runs grim per frame against the session's own desk and
costs more; the trade is a video of the thing being demonstrated.
The cursor is **not** in the frames, and that is not a capture setting: the
virtual pointer has no cursor image for the compositor to draw. See
docs/TODO/cursor-visibility-in-captures.md. The scrolling is visible by what
moves, which is why the page is scrolled rather than merely pointed at.
./proof-video.py [--profile DIR] [--out FILE]
"""
from __future__ import annotations
import argparse
import json
import pathlib
import subprocess
import sys
import time
import urllib.parse
import urllib.request
CDP_PORT = 9344
CDP = f"http://127.0.0.1:{CDP_PORT}"
SEARCH = "https://www.gunpost.ca/ads?key=SKS&sort=created&order=desc"
LISTINGS = [
"https://www.gunpost.ca/firearms/rifles/hamilton/sks-1951-russian-excellent-rifle-762x39-100000-obo",
"https://www.gunpost.ca/firearms/rifles/pembroke/1953-tula-russian-sks-762x39-crisp-markings-solid-hardwood-stock-box",
"https://www.gunpost.ca/firearms/rifles/markham-york-region/chinese-type-56-sks-hush-folding-stock-original-wood-stock",
]
MESSAGE = "hahaha I can see the world now"
#: Absolute, because the broker is a systemd unit and a unit's PATH has no system
#: profile in it: a bare `cowsay` fails inside foot with "No such file or directory".
COWSAY = "/nix/store/s4hlzmy3ll1bphdsy8vjdpcw447vbr93-cowsay-3.8.4/bin/cowsay"
#: How long each film runs. The beats below are timed to fit inside these.
BROWSER_SECONDS = 44
COW_SECONDS = 12
def call(path: str, method: str = "GET") -> dict:
req = urllib.request.Request(CDP + path, method=method)
with urllib.request.urlopen(req, timeout=15) as response:
body = response.read().decode()
return json.loads(body) if body.strip() else {}
def open_tab(url: str) -> str:
return call("/json/new?" + urllib.parse.quote(url, safe=""), method="PUT")["id"]
def activate(tab_id: str) -> None:
"""Bring a tab to the front. Cosmetic: never fail the recording over it."""
try:
call("/json/activate/" + tab_id)
except Exception as exc: # noqa: BLE001 - a failed activation is not fatal
print(f" (activate {tab_id[:8]} said: {exc})", flush=True)
def hhstream(*args: str, check: bool = False) -> str:
out = subprocess.run(["hhstream", *args], capture_output=True, text=True, timeout=300)
if out.returncode != 0:
print(f" (hhstream {' '.join(args[:3])} -> {out.returncode}: {out.stderr.strip()[:160]})")
if check:
raise SystemExit(f"hhstream {' '.join(args[:3])} failed")
return out.stdout.strip()
def desk_of(session: str) -> str:
"""The output name of a session's desk, as the broker recorded it."""
reply = json.loads(hhstream("session", "container", "--json") or "{}")
rows = reply.get("containers") or reply.get("sessions") or []
for row in rows:
if row.get("id") == session:
desk = row.get("output") or (row.get("container") or {}).get("output")
if desk:
return desk
raise SystemExit(f"{session} has no desk output: {json.dumps(rows)[:300]}")
def busy_work(session: str) -> None:
"""The pointer beats: scroll the results, then walk the listings.
Scrolling is the point of these beats. The pointer is aimed into this
session's own desk, so the scroll lands on its page and nowhere else -- and
because the cursor is not drawn, what a viewer sees is the page moving.
"""
lease = json.loads(hhstream("session", "pointer", "acquire", session, "--json") or "{}")
lease_id = lease.get("lease_id")
if not lease_id:
print(f" (no lease: {json.dumps(lease)[:200]})", flush=True)
return
# Down, down again, then back up: two directions, so it cannot be mistaken
# for a page that simply loaded a different part of itself.
hhstream("session", "pointer", "move", session, "--lease", lease_id, "--to", "0.5,0.55")
for amount, pause in ((6, 2.5), (6, 2.5), (-4, 2.0)):
print(f" scroll vertical {amount:+d}", flush=True)
hhstream(
"session", "pointer", "scroll", session,
"--lease", lease_id, "--axis", "vertical", "--amount", str(amount),
)
time.sleep(pause)
hhstream("session", "pointer", "release", session, "--lease", lease_id)
def film(session: str, seconds: int, out: pathlib.Path) -> None:
"""Record one desk until `seconds` have passed, then wait for the encoder."""
hhstream(
"record", "--output", desk_of(session), "--fps", "12",
"--seconds", str(seconds), "--out", str(out),
)
def main() -> int:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--profile", default="/var/lib/hermes/agent-eyes/gunpost-profile")
parser.add_argument("--out", default="/var/lib/hermes/agent-eyes/desks-proof.mp4")
parser.add_argument("--work-dir", default="/var/lib/hermes/agent-eyes/takes")
args = parser.parse_args()
out = pathlib.Path(args.out)
work = pathlib.Path(args.work_dir)
work.mkdir(parents=True, exist_ok=True)
pathlib.Path(args.profile).mkdir(parents=True, exist_ok=True)
a, b = work / "desk-a.mp4", work / "desk-b.mp4"
print("=== starting the browser on its own desk ===", flush=True)
browser = hhstream(
"session", "start", "chromium", "--",
"--ozone-platform=wayland", f"--user-data-dir={args.profile}",
"--no-first-run", "--no-default-browser-check", "--window-size=1280,720",
f"--remote-debugging-port={CDP_PORT}", "--remote-debugging-address=127.0.0.1",
"--new-window", "about:blank",
check=True,
)
print(f" {browser} on {desk_of(browser)}", flush=True)
time.sleep(4) # let the window map and the debugging port come up
print("=== film 1: the search, scrolling, and three listings ===", flush=True)
recording = subprocess.Popen(
["hhstream", "record", "--output", desk_of(browser), "--fps", "12",
"--seconds", str(BROWSER_SECONDS), "--out", str(a)],
stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
)
time.sleep(2)
search_tab = open_tab(SEARCH)
activate(search_tab)
print(" the SKS search is loading (Cloudflare included)", flush=True)
time.sleep(13)
busy_work(browser)
for n, url in enumerate(LISTINGS, 1):
activate(open_tab(url))
print(f" listing {n} in its own tab", flush=True)
time.sleep(4)
activate(search_tab)
time.sleep(2)
recording.wait(timeout=300)
print(f" {a}: {a.stat().st_size} bytes", flush=True)
print("=== clearing the browser away ===", flush=True)
print(f" {hhstream('session', 'kill', browser)[:90]}", flush=True)
time.sleep(2)
print("=== film 2: a terminal on its own desk, running cowsay ===", flush=True)
terminal = hhstream(
"session", "start", "foot", "--",
"-H", "-f", "monospace:size=20", COWSAY, MESSAGE,
check=True,
)
print(f" {terminal} on {desk_of(terminal)}", flush=True)
time.sleep(2)
film(terminal, COW_SECONDS, b)
print(f" {b}: {b.stat().st_size} bytes", flush=True)
print(f" {hhstream('session', 'kill', terminal)[:90]}", flush=True)
print("=== joining the two films ===", flush=True)
listing = work / "concat.txt"
listing.write_text(f"file '{a}'\nfile '{b}'\n")
joined = subprocess.run(
["ffmpeg", "-v", "error", "-f", "concat", "-safe", "0", "-i", str(listing),
"-c", "copy", "-y", str(out)],
capture_output=True, text=True,
)
if joined.returncode != 0:
raise SystemExit(f"join failed: {joined.stderr.strip()[:300]}")
probe = subprocess.run(
["ffprobe", "-v", "error", "-show_entries", "format=duration,size",
"-of", "default=noprint_wrappers=1", str(out)],
capture_output=True, text=True,
)
print(f" {out}\n {probe.stdout.strip().replace(chr(10), ' ')}", flush=True)
return 0
if __name__ == "__main__":
sys.exit(main())
+12 -2
View File
@@ -79,8 +79,18 @@ in ''
} }
input { input {
# No pointer will ever exist; do not let focus depend on one. # Focus follows the pointer. This was `0`, on the reasoning written here
follow_mouse = 0 # that no pointer would ever exist — true when it was written, and false
# now that the broker injects one.
#
# It is not a preference. With `0` Hyprland delivers pointer events only to
# the keyboard-focused window: measured in `vm-pointer`, a click aimed at
# 0.25,0.75 of a session's container placed the cursor at x=299 y=519,
# inside that window, and **both** clients reported nothing — the click was
# aimed correctly and delivered nowhere. The cursor lease already
# serialises input, so letting the pointer move keyboard focus cannot make
# two sessions fight over the keyboard.
follow_mouse = 1
} }
# Frames cost capture bandwidth and CPU, and no one is watching the easing. # Frames cost capture bandwidth and CPU, and no one is watching the easing.
+1 -1
View File
@@ -27,7 +27,7 @@ DEFAULT_SOCKET = "/run/hypr-stream/stream.sock"
#: Contract: the published socket is mode 0660 (ownership is the unit's job). #: Contract: the published socket is mode 0660 (ownership is the unit's job).
SOCKET_MODE = 0o660 SOCKET_MODE = 0o660
#: Contract default for `services.hyprland-headless-stream.captureCommand`. #: Contract default for `services.hyprland-headless-stream.captureCommand`.
DEFAULT_CAPTURE_COMMAND = ["grim", "-t", "png", "-"] DEFAULT_CAPTURE_COMMAND = ["grim", "-c", "-t", "png", "-"]
_BACKLOG = 64 _BACKLOG = 64
_WAIT_TICK = 0.2 # how often a blocked client thread re-checks for shutdown _WAIT_TICK = 0.2 # how often a blocked client thread re-checks for shutdown
+4
View File
@@ -193,6 +193,10 @@ class WaylandSink:
self._pointer(output).button(_now_ms(), code, state) self._pointer(output).button(_now_ms(), code, state)
self._flush() self._flush()
def axis(self, output: str, axis: int, value: float) -> None:
self._pointer(output).axis(_now_ms(), axis, value)
self._flush()
def frame(self, output: str) -> None: def frame(self, output: str) -> None:
self._pointer(output).frame() self._pointer(output).frame()
self._flush() self._flush()
+23
View File
@@ -27,7 +27,30 @@ CODES: dict[str, int] = {
#: ``wl_pointer.button_state``. Named on the wire so a log line reads. #: ``wl_pointer.button_state``. Named on the wire so a log line reads.
STATES: dict[str, int] = {"up": 0, "down": 1} STATES: dict[str, int] = {"up": 0, "down": 1}
#: ``wl_pointer_axis`` values. Named "vertical"/"horizontal" rather than
#: up/down/left/right, because the sign is what carries the direction: the
#: protocol's positive vertical scroll is *down*, and a caller asking for
#: "down" while a negative value means down is how a scroll ends up backwards.
AXES: dict[str, int] = {"vertical": 0, "horizontal": 1}
#: The largest scroll a single request may carry, in wheel clicks. Bounded so a
#: decimal-point slip (3 -> 30000) is a refusal rather than a session scrolled
#: into next week, where the only clue is that the page will not move back.
MAX_AXIS_VALUE = 100.0
NAMES = tuple(CODES) NAMES = tuple(CODES)
AXIS_NAMES = tuple(AXES)
def axis_for(name: str) -> int:
"""The protocol axis for a name, or :class:`ValueError` naming the set."""
try:
return AXES[name]
except KeyError:
raise ValueError(
f"unknown axis {name!r} (known: {', '.join(AXIS_NAMES)})"
) from None
def code_for(name: str) -> int: def code_for(name: str) -> int:
+39
View File
@@ -39,6 +39,13 @@ class PointerSink(Protocol):
def button(self, output: str, code: int, state: int) -> None: def button(self, output: str, code: int, state: int) -> None:
"""An evdev button code and a ``wl_pointer.button_state``.""" """An evdev button code and a ``wl_pointer.button_state``."""
def axis(self, output: str, axis: int, value: float) -> None:
"""Scroll by ``value`` wheel clicks along ``axis`` (a ``wl_pointer_axis``).
Positive vertical is *down* and positive horizontal is *right*, which is
the protocol's direction and not a choice of ours.
"""
def frame(self, output: str) -> None: def frame(self, output: str) -> None:
"""Commit the queued events. Nothing is delivered until this.""" """Commit the queued events. Nothing is delivered until this."""
@@ -107,6 +114,37 @@ class Dispatcher:
self._pressed.pop(name, None) self._pressed.pop(name, None)
return {"ok": True} return {"ok": True}
def _op_axis(self, request: dict) -> dict[str, Any]:
"""Scroll the aimed output.
The output is implicit, like ``button``: there is one cursor, and a
scroll belongs to whatever it is currently over. A caller that wants to
scroll somewhere else warps first, which is also what makes the gesture
serialisable by the lease.
"""
name = request.get("axis", "vertical")
if not isinstance(name, str):
raise Rejected("'axis' must be a string")
try:
code = buttons.axis_for(name)
except ValueError as exc:
raise Rejected(str(exc)) from None
value = request.get("value", 1.0)
if isinstance(value, bool) or not isinstance(value, (int, float)):
raise Rejected("'value' must be a number of wheel clicks")
value = float(value)
if value != value or value in (float("inf"), float("-inf")):
raise Rejected("'value' must be finite")
if abs(value) > buttons.MAX_AXIS_VALUE:
raise Rejected(
f"'value' {value} exceeds the {buttons.MAX_AXIS_VALUE:g} wheel-click limit"
)
self._sink.axis(self._aimed(), code, value)
return {"ok": True}
def _op_frame(self, request: dict) -> dict[str, Any]: def _op_frame(self, request: dict) -> dict[str, Any]:
self._sink.frame(self._aimed()) self._sink.frame(self._aimed())
return {"ok": True} return {"ok": True}
@@ -131,6 +169,7 @@ class Dispatcher:
_ops = { _ops = {
"warp": _op_warp, "warp": _op_warp,
"button": _op_button, "button": _op_button,
"axis": _op_axis,
"frame": _op_frame, "frame": _op_frame,
"pressed": _op_pressed, "pressed": _op_pressed,
"release_all": _op_release_all, "release_all": _op_release_all,
+54
View File
@@ -302,6 +302,14 @@ def _build_namespace(parser: argparse.ArgumentParser) -> None:
#: and this list only buys a typo an exit 2 instead of an exit 10. #: and this list only buys a typo an exit 2 instead of an exit 10.
BUTTONS = ("left", "right", "middle", "side", "extra", "forward", "back", "task") BUTTONS = ("left", "right", "middle", "side", "extra", "forward", "back", "task")
#: Axis names the driver knows (``hhstream_pointer.buttons.AXES``), mirrored for
#: the same reason and with the same caveat: the driver remains the authority.
#: Direction is the sign of the amount, so there is deliberately no "down" or
#: "up" here -- the protocol fixes positive as down (vertical) and right
#: (horizontal), and a second vocabulary over it is how a scroll ends up
#: backwards.
AXES = ("vertical", "horizontal")
def _build_pointer(verbs: Any, common: argparse.ArgumentParser) -> None: def _build_pointer(verbs: Any, common: argparse.ArgumentParser) -> None:
"""``session pointer <verb>`` — the one cursor, under a lease (PLAN §6). """``session pointer <verb>`` — the one cursor, under a lease (PLAN §6).
@@ -409,6 +417,29 @@ def _build_pointer(verbs: Any, common: argparse.ArgumentParser) -> None:
"--button", default="left", choices=BUTTONS, metavar="B", help=_button_help() "--button", default="left", choices=BUTTONS, metavar="B", help=_button_help()
) )
scroll = with_lease(add_gesture(
"scroll",
_cmd_pointer_scroll,
help="scroll the session's screen where the cursor is",
description=(
"Scroll the session's screen. --amount is signed wheel clicks: "
"positive scrolls down (vertical) or right (horizontal), which is "
"the protocol's direction rather than a choice of ours. The cursor "
"does not move, so the resume point keeps its coordinates -- but it "
"is re-stamped, because a scroll is the session acting and a later "
"resume should know when it last did."
),
))
scroll.add_argument("id", help="session id")
scroll.add_argument(
"--axis", default="vertical", choices=AXES, metavar="A",
help="vertical (default) or horizontal",
)
scroll.add_argument(
"--amount", type=float, default=3.0, metavar="N",
help="wheel clicks, signed; positive is down/right (default: 3)",
)
for name, state in (("down", "down"), ("up", "up")): for name, state in (("down", "down"), ("up", "up")):
sub = with_lease(add_gesture( sub = with_lease(add_gesture(
name, name,
@@ -951,6 +982,29 @@ def _cmd_pointer_click(args: argparse.Namespace) -> int:
return _finish_gesture(args, reply, render, args.id) return _finish_gesture(args, reply, render, args.id)
@_guard
def _cmd_pointer_scroll(args: argparse.Namespace) -> int:
lease_id = _lease_of(args, args.id)
if not lease_id:
return _no_lease(args, args.id)
reply = _request(
args,
"pointer.scroll",
id=args.id,
lease_id=lease_id,
axis=args.axis,
amount=args.amount,
)
def render(r: dict[str, Any]) -> str:
return (
f"{r.get('id')}: scrolled {r.get('axis')} by {r.get('amount')} "
f"at {fmt.render_bookmark(r.get('bookmark'))}"
)
return _finish_gesture(args, reply, render, args.id)
@_guard @_guard
def _cmd_pointer_button(args: argparse.Namespace) -> int: def _cmd_pointer_button(args: argparse.Namespace) -> int:
lease_id = _lease_of(args, args.id) lease_id = _lease_of(args, args.id)
+12 -1
View File
@@ -23,7 +23,7 @@ from .clock import Clock, parse_duration
from .compositor import CompositorError, CompositorTarget from .compositor import CompositorError, CompositorTarget
from .launcher import Launcher from .launcher import Launcher
from .manager import DEFAULT_SETTLE_S, ManagerConfig, SessionManager from .manager import DEFAULT_SETTLE_S, ManagerConfig, SessionManager
from .policy import Policy, PolicyError, load_policy from .policy import Policy, PolicyError, load_policy, load_settings
from .registry import Registry from .registry import Registry
from .server import ControlServer from .server import ControlServer
@@ -166,12 +166,23 @@ async def serve(
# against whatever directory the app happens to be started in. # against whatever directory the app happens to be started in.
state_dir = Path(args.state_dir).resolve() state_dir = Path(args.state_dir).resolve()
runtime_dir = Path(args.runtime_dir).resolve() runtime_dir = Path(args.runtime_dir).resolve()
# The same file the allowlist comes from. Read here rather than in the
# manager so the manager keeps taking a plain ManagerConfig, which is what
# its tests build directly.
settings = load_settings(args.config)
config = ManagerConfig( config = ManagerConfig(
state_dir=state_dir, state_dir=state_dir,
runtime_dir=runtime_dir, runtime_dir=runtime_dir,
policy=policy, policy=policy,
settle_s=args.settle, settle_s=args.settle,
compositor=compositor, compositor=compositor,
containers_enabled=settings.containers_enabled,
container_base=settings.container_base,
container_slots=settings.container_slots,
container_mode=settings.container_mode,
quarantine_workspace=settings.quarantine_workspace,
pointer_ttl_s=settings.pointer_ttl_s,
pointer_max_s=settings.pointer_max_s,
) )
manager = SessionManager( manager = SessionManager(
config, config,
+64
View File
@@ -86,9 +86,15 @@ class ContainerHost:
mode: tuple[int, int] = DEFAULT_MODE, mode: tuple[int, int] = DEFAULT_MODE,
quarantine_workspace: str = DEFAULT_QUARANTINE_WORKSPACE, quarantine_workspace: str = DEFAULT_QUARANTINE_WORKSPACE,
ipc: HyprlandIPC | None = None, ipc: HyprlandIPC | None = None,
enabled: bool = True,
) -> None: ) -> None:
self.runtime_dir = None if runtime_dir is None else Path(runtime_dir) self.runtime_dir = None if runtime_dir is None else Path(runtime_dir)
self.allocator = allocator self.allocator = allocator
#: Whether this broker gives sessions containers at all. Off is a
#: configuration, not a failure: `available()` reports it, and the manager
#: already knows what to do with an unavailable host (sessions without
#: containers), so nothing else has to branch on it.
self.enabled = enabled
self.mode = mode self.mode = mode
self.quarantine_workspace = quarantine_workspace self.quarantine_workspace = quarantine_workspace
self._wanted_signature = signature self._wanted_signature = signature
@@ -138,8 +144,19 @@ class ContainerHost:
``version`` is the cheapest round trip that proves the socket is a live ``version`` is the cheapest round trip that proves the socket is a live
compositor and not a leftover inode. compositor and not a leftover inode.
A host switched off by configuration is unavailable on purpose, and the
reason says which key turned it off -- otherwise a machine with
containers disabled looks identical to one whose compositor is down, and
the operator goes hunting for a broken compositor that is fine.
""" """
if not self.enabled:
self._last_error = (
"containers are disabled in the config "
"([compositor] containers_enabled = false)"
)
return False
try: try:
self.ipc().version() self.ipc().version()
except CompositorUnavailable as exc: except CompositorUnavailable as exc:
@@ -219,6 +236,53 @@ class ContainerHost:
raise raise
return plan return plan
def claim(
self, app_id: str, pid: int | None, workspace: int, output: str | None
) -> ContainerFacts | None:
"""Put this session's window on its desk's workspace, and show it.
Assignment is by **pid**, not by class. The compositor reports every
client's pid and the broker knows each session's, so this needs nothing
from the application -- which is the point: the per-session window rule
matches a class stamp that foot, GTK, Qt and the chromium wrapper all
ignore, so a rule-bound desk is empty while the window it was made for
sits in the quarantine backstop.
Two dispatches, in this order: move the window, then make its workspace
the one its own output is showing. Moving alone leaves the window on a
workspace nobody is looking at -- ``foreground`` false, and a frame
capture of wallpaper.
Everything is read back. What the caller records is where the window
*is* afterwards, never where it was sent.
"""
facts = self.read_back(app_id, pid)
if facts is None:
return None
window = facts.window or {}
address = window.get("address")
if facts.workspace != workspace and address:
self.dispatch(
[
"hyprctl",
"dispatch",
"movetoworkspacesilent",
f"{workspace},address:{address}",
]
)
if output is not None and not self.is_showing(workspace, output):
self.dispatch(["hyprctl", "dispatch", "workspace", str(workspace)])
return self.read_back(app_id, pid) or facts
def is_showing(self, workspace: int, output: str) -> bool:
"""Is ``output`` the one displaying ``workspace`` right now?"""
for monitor in self.ipc().monitors():
if monitor.name == output:
return monitor.active_workspace_id == workspace
return False
def remove(self, output: str) -> None: def remove(self, output: str) -> None:
"""Remove one output of ours. Refuses a name we could not have made.""" """Remove one output of ours. Refuses a name we could not have made."""
+11
View File
@@ -83,6 +83,17 @@ class PointerDriver:
def button(self, button: str, state: str) -> None: def button(self, button: str, state: str) -> None:
self.request({"op": "button", "button": button, "state": state}) self.request({"op": "button", "button": button, "state": state})
def axis(self, axis_name: str, value: float) -> None:
"""Scroll the aimed output by ``value`` wheel clicks along ``axis_name``.
Positive vertical is down, positive horizontal is right (the protocol's
direction). The output is whatever the last ``warp`` aimed at, so a
scroll into another session is a warp first -- which is also what the
lease serialises.
"""
self.request({"op": "axis", "axis": axis_name, "value": float(value)})
def frame(self) -> None: def frame(self) -> None:
"""Commit what was queued. Nothing the driver was sent is delivered """Commit what was queued. Nothing the driver was sent is delivered
to the compositor until this, so every gesture ends with one.""" to the compositor until this, so every gesture ends with one."""
+80 -6
View File
@@ -33,6 +33,7 @@ watching an idle desktop.
from __future__ import annotations from __future__ import annotations
import json import json
import logging
import os import os
import socket import socket
import stat import stat
@@ -49,6 +50,8 @@ INSTANCE_SUBDIR = "hypr"
REQUEST_SOCKET_NAME = ".socket.sock" REQUEST_SOCKET_NAME = ".socket.sock"
EVENT_SOCKET_NAME = ".socket2.sock" EVENT_SOCKET_NAME = ".socket2.sock"
log = logging.getLogger("hhstreamd.ipc")
#: Requests are local, synchronous and tiny. The compositor answers at once or #: Requests are local, synchronous and tiny. The compositor answers at once or
#: it is wedged; this only bounds the pathological case where it accepts the #: it is wedged; this only bounds the pathological case where it accepts the
#: connection and then stops. It is a class attribute so a test can shorten it #: connection and then stops. It is a class attribute so a test can shorten it
@@ -392,9 +395,23 @@ class HyprlandIPC:
raise CompositorUnavailable(f"compositor request socket {request} is not a socket") raise CompositorUnavailable(f"compositor request socket {request} is not a socket")
return cls(request, directory / EVENT_SOCKET_NAME) return cls(request, directory / EVENT_SOCKET_NAME)
@staticmethod @classmethod
def _sole_instance(hypr: Path) -> str: def _sole_instance(cls, hypr: Path) -> str:
"""The one directory under ``<runtime_dir>/hypr``, or a refusal.""" """The live instance under ``<runtime_dir>/hypr``, or a refusal.
A restart leaves its directory behind — the compositor's log outlives
it — so after two restarts this directory holds two names and neither is
obviously the real one. Refusing outright (which this used to do) is
safe but wrong after every rebuild: the broker cannot reach any
compositor, so no container can be created and the pointer has nothing to
aim at, all for want of a corpse being cleared.
So each candidate is *asked*, on its own request socket, for its version,
and one that answers is the compositor. Nothing is inferred from
timestamps unless more than one answers — a dead instance's log file can
easily be newer than a live instance's socket, so age is the weakest
signal available and is used only to break a tie.
"""
try: try:
with os.scandir(hypr) as entries: with os.scandir(hypr) as entries:
@@ -407,12 +424,69 @@ class HyprlandIPC:
raise CompositorUnavailable( raise CompositorUnavailable(
f"compositor instance dir {hypr} holds no instance; is the compositor running?" f"compositor instance dir {hypr} holds no instance; is the compositor running?"
) )
if len(names) > 1: if len(names) == 1:
return names[0]
answered = [name for name in names if cls._answers(hypr / name)]
if len(answered) == 1:
log.info(
"compositor instance dir %s holds %d instances (%s); %s answered, "
"using it and ignoring the rest",
hypr,
len(names),
", ".join(names),
answered[0],
)
return answered[0]
if not answered:
raise CompositorUnavailable( raise CompositorUnavailable(
f"compositor instance dir {hypr} holds {len(names)} instances " f"compositor instance dir {hypr} holds {len(names)} instances "
f"({', '.join(names)}); set {SIGNATURE_ENV} or pass a signature" f"({', '.join(names)}) and none of them answered; set "
f"{SIGNATURE_ENV} or pass a signature"
) )
return names[0] newest = max(answered, key=lambda name: cls._socket_age(hypr / name))
log.warning(
"compositor instance dir %s holds %d answering instances (%s); using the "
"most recently modified, %s — pass %s to be certain",
hypr,
len(answered),
", ".join(answered),
newest,
SIGNATURE_ENV,
)
return newest
@classmethod
def _answers(cls, directory: Path) -> bool:
"""Does this instance dir hold a request socket that replies?
The probe is the real protocol, not a file check: a socket that exists
but belongs to a compositor that is gone looks identical on disk, and
the only way to tell them apart is to ask a question with a known
answer.
"""
request = directory / REQUEST_SOCKET_NAME
try:
mode = request.stat().st_mode
except OSError:
return False
if not stat.S_ISSOCK(mode):
return False
try:
cls(request, directory / EVENT_SOCKET_NAME).query("version")
except (CompositorUnavailable, OSError):
return False
return True
@staticmethod
def _socket_age(directory: Path) -> float:
"""Modification time of the request socket, for tie-breaking only."""
try:
return (directory / REQUEST_SOCKET_NAME).stat().st_mtime
except OSError:
return 0.0
# -- the request socket ------------------------------------------------ # -- the request socket ------------------------------------------------
+136 -1
View File
@@ -50,7 +50,12 @@ from . import containers, procmon
from .audit import AuditLog from .audit import AuditLog
from .clock import Clock, iso from .clock import Clock, iso
from .compositor import CompositorTarget from .compositor import CompositorTarget
from .container_host import DEFAULT_MODE, DEFAULT_QUARANTINE_WORKSPACE, ContainerHost from .container_host import (
DEFAULT_MODE,
DEFAULT_QUARANTINE_WORKSPACE,
ContainerFacts,
ContainerHost,
)
from .driver import DriverRejected, DriverUnavailable, PointerDriver from .driver import DriverRejected, DriverUnavailable, PointerDriver
from .ipc import CompositorUnavailable from .ipc import CompositorUnavailable
from .launcher import Launcher from .launcher import Launcher
@@ -116,6 +121,12 @@ DEFAULT_CONTAINER_SLOTS = 4
#: and "click 10000 times" is not a gesture. #: and "click 10000 times" is not a gesture.
MAX_CLICK_COUNT = 3 MAX_CLICK_COUNT = 3
#: The largest scroll one request may carry, in wheel clicks. The driver
#: enforces the same ceiling with its own copy (they are separate processes and
#: neither imports the other); this one exists so the broker refuses early with a
#: message naming the caller's value rather than relaying the driver's.
MAX_SCROLL_AMOUNT = 100.0
SIGNALS_BY_NAME = { SIGNALS_BY_NAME = {
"TERM": signal.SIGTERM, "TERM": signal.SIGTERM,
"SIGTERM": signal.SIGTERM, "SIGTERM": signal.SIGTERM,
@@ -161,6 +172,11 @@ class ManagerConfig:
signature: str | None = None signature: str | None = None
container_base: int = DEFAULT_CONTAINER_BASE container_base: int = DEFAULT_CONTAINER_BASE
container_slots: int = DEFAULT_CONTAINER_SLOTS container_slots: int = DEFAULT_CONTAINER_SLOTS
#: Whether sessions get their own output and workspace. False means "process
#: groups on the shared output", which is what the machine ran before
#: containers existed, and what a deployment that cannot match windows to
#: containers yet should be configured with.
containers_enabled: bool = True
container_mode: tuple[int, int] = DEFAULT_MODE container_mode: tuple[int, int] = DEFAULT_MODE
quarantine_workspace: str = DEFAULT_QUARANTINE_WORKSPACE quarantine_workspace: str = DEFAULT_QUARANTINE_WORKSPACE
readback_s: float = DEFAULT_READBACK_S readback_s: float = DEFAULT_READBACK_S
@@ -213,6 +229,7 @@ class SessionManager:
signature=config.signature, signature=config.signature,
mode=config.container_mode, mode=config.container_mode,
quarantine_workspace=config.quarantine_workspace, quarantine_workspace=config.quarantine_workspace,
enabled=config.containers_enabled,
) )
# The lease runs on CLOCK_BOOTTIME like every other deadline here, so a # The lease runs on CLOCK_BOOTTIME like every other deadline here, so a
# host suspended mid-gesture does not come back believing the lease is # host suspended mid-gesture does not come back believing the lease is
@@ -529,6 +546,10 @@ class SessionManager:
# `container.reconcile` picks it up later. Nothing is invented. # `container.reconcile` picks it up later. Nothing is invented.
log.info("%s: no window in clients -j yet; workspace and window stay null", session.id) log.info("%s: no window in clients -j yet; workspace and window stay null", session.id)
return return
scene = session.container or {}
workspace, output = scene.get("workspace"), scene.get("output")
if workspace is not None and output is not None:
facts = await self._claim(session, app_id, facts, workspace, output)
session.workspace = facts.workspace session.workspace = facts.workspace
session.window = facts.window session.window = facts.window
session.foreground = facts.foreground session.foreground = facts.foreground
@@ -551,6 +572,44 @@ class SessionManager:
) )
self.registry.persist(session, force=True) self.registry.persist(session, force=True)
async def _claim(
self,
session: Session,
app_id: str,
facts: ContainerFacts,
workspace: int,
output: str,
) -> ContainerFacts:
"""Move the session's window onto its desk, if it is not already there.
Skipped when the window is already home *and* the desk is showing it,
so the steady state costs nothing. Otherwise the compositor is asked to
move it and then read back, and what gets recorded is where the window
ended up -- an audit event is written only when something actually
changed, so the trail stays a record of moves rather than of polls.
"""
if facts.workspace == workspace and facts.foreground:
return facts
try:
claimed = self.host.claim(app_id, session.pid, workspace, output)
except CompositorUnavailable as exc:
log.error("%s: claiming its window failed: %s", session.id, exc)
self.containers_available = False
return facts
if claimed is None:
return facts
if claimed.workspace != facts.workspace or claimed.foreground != facts.foreground:
self.audit.emit(
"container_claimed",
session=session.id,
from_workspace=facts.workspace,
workspace=claimed.workspace,
foreground=claimed.foreground,
address=(claimed.window or {}).get("address"),
)
return claimed
async def shutdown(self) -> None: async def shutdown(self) -> None:
"""Stop watching, flush the mirror. **Sessions are left running.** """Stop watching, flush the mirror. **Sessions are left running.**
@@ -1479,6 +1538,15 @@ class SessionManager:
except CompositorUnavailable as exc: except CompositorUnavailable as exc:
self.containers_available = False self.containers_available = False
return [{"id": session.id, "field": "compositor", "error": str(exc)}] return [{"id": session.id, "field": "compositor", "error": str(exc)}]
scene = session.container or {}
workspace, output = scene.get("workspace"), scene.get("output")
if facts is not None and workspace is not None and output is not None:
# Reconcile is the second place a window can be found: the first is
# the read-back at start, and an application slow to map one -- or
# one that reopened -- turns up only here. Claiming it is a repair,
# so it happens before the diff is taken: what gets reported is the
# difference between the old record and where the window now is.
facts = await self._claim(session, app_id, facts, workspace, output)
observed = { observed = {
"workspace": None if facts is None else facts.workspace, "workspace": None if facts is None else facts.workspace,
"window": None if facts is None else facts.window, "window": None if facts is None else facts.window,
@@ -1652,6 +1720,44 @@ class SessionManager:
"bookmark": self._restore_point(session, output), "bookmark": self._restore_point(session, output),
} }
async def pointer_scroll(
self, session_id: str, lease_id: str, axis: Any = "vertical", amount: Any = 3
) -> dict[str, Any]:
"""Scroll the session's screen.
Same shape as a click -- the lease is checked first, then the driver is
told, then the resume point is re-stamped -- with one difference that is
worth being precise about: **the cursor does not move**. The
coordinates are therefore re-affirmed rather than changed. The stamp
advances, because a scroll is the session acting and a later resume
should know when it last did -- but not one pixel of the position
changes, and a bookmark that claimed otherwise would be fiction.
"""
session, output = await self._refresh(session_id, lease_id)
name = _axis_name(axis if axis is not None else "vertical")
clicks = _checked_amount(amount)
await self._drive("axis", name, clicks)
await self._drive("frame")
# A bookmark from a container this session no longer has is not a
# position in this one, so a missing or stale one falls back to the
# centre that the acquire would have aimed at.
known = self._restore_point(session, output) or {}
mark = self._mark(
session,
lease_id,
output,
float(known.get("x", 0.5)),
float(known.get("y", 0.5)),
)
self.audit.emit("pointer_scroll", session=session.id, axis=name, amount=clicks)
return {
"id": session.id,
"axis": name,
"amount": clicks,
"bookmark": mark,
}
async def pointer_release(self, session_id: str, lease_id: str) -> dict[str, Any]: async def pointer_release(self, session_id: str, lease_id: str) -> dict[str, Any]:
"""Give the cursor back, leaving nothing pressed and a bookmark behind.""" """Give the cursor back, leaving nothing pressed and a bookmark behind."""
@@ -2073,6 +2179,35 @@ def _button_name(value: Any) -> str:
return value return value
def _axis_name(value: Any) -> str:
if not isinstance(value, str) or not value:
raise ManagerError("invalid_argument", f"axis must be a name, got {value!r}")
return value
def _checked_amount(value: Any) -> float:
"""A scroll amount in wheel clicks, signed.
Signed rather than split into `--up`/`--down`: the protocol carries direction
in the sign (positive vertical is down), and a second vocabulary on top of it
is how a scroll ends up backwards. Zero is refused because it is never
intended -- it is a caller that forgot to pass an amount.
"""
if isinstance(value, bool) or not isinstance(value, (int, float)):
raise ManagerError(
"invalid_argument", f"amount must be a number of wheel clicks, got {value!r}"
)
amount = float(value)
if amount == 0 or abs(amount) > MAX_SCROLL_AMOUNT:
raise ManagerError(
"invalid_argument",
f"amount must be non-zero and at most {MAX_SCROLL_AMOUNT:g} wheel clicks, "
f"got {amount}",
)
return amount
def _checked_count(value: Any) -> int: def _checked_count(value: Any) -> int:
if isinstance(value, bool) or not isinstance(value, int): if isinstance(value, bool) or not isinstance(value, int):
raise ManagerError("invalid_argument", f"count must be an integer, got {value!r}") raise ManagerError("invalid_argument", f"count must be an integer, got {value!r}")
+88
View File
@@ -169,6 +169,94 @@ def check_argv(argv: Any) -> None:
# -------------------------------------------------------------------------- # --------------------------------------------------------------------------
@dataclass(frozen=True, slots=True)
class Settings:
"""What the config file says that is not the allowlist.
Kept beside :class:`Policy` because both come out of the same TOML file, and
read by ``__main__`` rather than by the manager, so the manager keeps taking a
plain :class:`~hhstreamd.manager.ManagerConfig` that a test can build
directly. Before this existed the module rendered these keys into the file
and the broker read none of them: a deployment could turn containers off, or
move a lease deadline, and be silently ignored while the defaults happened to
make it look honoured.
The defaults mirror the manager's own; the module always renders every key,
so they matter only for a config written by hand.
"""
containers_enabled: bool = True
container_base: int = 1000
container_slots: int = 4
container_mode: tuple[int, int] = (1280, 720)
quarantine_workspace: str = "special:quarantine"
pointer_ttl_s: float = 30.0
pointer_max_s: float = 300.0
def _mode(value: Any, key: str) -> tuple[int, int]:
"""Parse ``"1280x720"`` into ``(1280, 720)``; refuse anything else.
Refusing matters more than being permissive here: a container created at the
wrong size is a session whose window does not fit its own screen, which looks
like an application bug rather than a typo in a config file.
"""
if isinstance(value, str) and "x" in value:
width, _, height = value.partition("x")
try:
return int(width), int(height)
except ValueError:
pass
if isinstance(value, (list, tuple)) and len(value) == 2:
return int(value[0]), int(value[1])
raise PolicyError(
"invalid_argument",
f"[compositor] {key} must look like \"1280x720\", got {value!r}",
)
def settings_from_dict(data: dict[str, Any]) -> Settings:
defaults = data.get("defaults", {})
if not isinstance(defaults, dict):
raise PolicyError("invalid_argument", "[defaults] must be a table")
compositor = data.get("compositor", {})
if not isinstance(compositor, dict):
raise PolicyError("invalid_argument", "[compositor] must be a table")
pointer = data.get("pointer", {})
if not isinstance(pointer, dict):
raise PolicyError("invalid_argument", "[pointer] must be a table")
return Settings(
containers_enabled=bool(compositor.get("containers_enabled", True)),
container_base=int(compositor.get("workspace_base", 1000)),
container_slots=int(defaults.get("max_containers", 4)),
container_mode=_mode(compositor.get("container_mode", "1280x720"), "container_mode"),
quarantine_workspace=str(
compositor.get("quarantine_workspace", "special:quarantine")
),
pointer_ttl_s=float(pointer.get("lease_ttl_s", 30.0)),
pointer_max_s=float(pointer.get("lease_max_s", 300.0)),
)
def load_settings(path: Path | str | None) -> Settings:
"""Read the non-allowlist settings from the file :func:`load_policy` reads.
A missing file is the defaults, for the same reason a missing allowlist is an
empty one: a misconfigured deployment should behave conservatively, not fail
to start.
"""
if path is None:
return Settings()
path = Path(path)
if not path.exists():
return Settings()
with open(path, "rb") as fh:
return settings_from_dict(tomllib.load(fh))
def load_policy(path: Path | str | None) -> Policy: def load_policy(path: Path | str | None) -> Policy:
"""Read the allowlist from a TOML file. """Read the allowlist from a TOML file.
+2
View File
@@ -148,6 +148,7 @@ VERBS = (
"pointer.move", "pointer.move",
"pointer.button", "pointer.button",
"pointer.click", "pointer.click",
"pointer.scroll",
"pointer.release", "pointer.release",
"pointer.state", "pointer.state",
) )
@@ -171,6 +172,7 @@ MUTATING_VERBS = frozenset(
"pointer.move", "pointer.move",
"pointer.button", "pointer.button",
"pointer.click", "pointer.click",
"pointer.scroll",
"pointer.release", "pointer.release",
} }
) )
+13
View File
@@ -70,6 +70,7 @@ class ControlServer:
"pointer.move": self._op_pointer_move, "pointer.move": self._op_pointer_move,
"pointer.button": self._op_pointer_button, "pointer.button": self._op_pointer_button,
"pointer.click": self._op_pointer_click, "pointer.click": self._op_pointer_click,
"pointer.scroll": self._op_pointer_scroll,
"pointer.release": self._op_pointer_release, "pointer.release": self._op_pointer_release,
"pointer.state": self._op_pointer_state, "pointer.state": self._op_pointer_state,
} }
@@ -356,6 +357,18 @@ class ControlServer:
) )
) )
async def _op_pointer_scroll(
self, request: dict[str, Any], caller: dict[str, Any]
) -> dict[str, Any]:
return protocol.ok(
**await self.manager.pointer_scroll(
request.get("id"),
request.get("lease_id"),
request.get("axis", "vertical"),
request.get("amount", 3),
)
)
async def _op_pointer_release( async def _op_pointer_release(
self, request: dict[str, Any], caller: dict[str, Any] self, request: dict[str, Any], caller: dict[str, Any]
) -> dict[str, Any]: ) -> dict[str, Any]:
+81 -5
View File
@@ -1,17 +1,25 @@
"""Finding the instance, and refusing to pick one when there is a choice. """Finding the instance: one answer beats a name that merely exists.
Two instances and no signature is the dangerous case: a broker that guessed Two instance dirs and no signature used to be a hard refusal. It still is when
would aim its queries — and later its pointer — at someone else's desktop. neither answers — a broker that guessed would aim its queries, and later its
pointer, at someone else's desktop. But a *dead* instance directory is left
behind by every compositor restart, so refusing on two names made the broker
blind after every rebuild. The compositor that answers a version request is the
one that exists; the corpse beside it is a log file.
""" """
from __future__ import annotations from __future__ import annotations
import os
import socket import socket
import threading
import pytest import pytest
from hhstreamd.ipc import SIGNATURE_ENV, CompositorUnavailable, HyprlandIPC from hhstreamd.ipc import SIGNATURE_ENV, CompositorUnavailable, HyprlandIPC
BANNER = b"Hyprland 0.56.2 built from branch v0.56.2-b\nTag: v0.56.2\n"
def bind(path): def bind(path):
"""A bound, listening socket at ``path``, so discovery sees a real one.""" """A bound, listening socket at ``path``, so discovery sees a real one."""
@@ -22,6 +30,33 @@ def bind(path):
return sock return sock
def answering(path):
"""A socket with a compositor behind it: it replies to a version request."""
sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
sock.bind(str(path))
sock.listen(4)
def serve() -> None:
while True:
try:
conn, _ = sock.accept()
except OSError:
return
with conn:
# The client half-closes after writing, so the request ends at
# EOF -- the same shape the real compositor sees.
while conn.recv(4096):
pass
try:
conn.sendall(BANNER)
except OSError:
pass
threading.Thread(target=serve, daemon=True).start()
return sock
def test_single_instance_is_discovered(runtime_dir, instance, monkeypatch): def test_single_instance_is_discovered(runtime_dir, instance, monkeypatch):
monkeypatch.delenv(SIGNATURE_ENV, raising=False) monkeypatch.delenv(SIGNATURE_ENV, raising=False)
sock = bind(instance / ".socket.sock") sock = bind(instance / ".socket.sock")
@@ -47,19 +82,60 @@ def test_zero_instances_under_hypr(runtime_dir, monkeypatch):
HyprlandIPC.discover(runtime_dir) HyprlandIPC.discover(runtime_dir)
def test_two_instances_without_a_signature_refuses(runtime_dir, instance, monkeypatch): def test_two_instances_and_one_answers_uses_the_one_that_answers(
runtime_dir, instance, monkeypatch
):
"""The corpse beside the live compositor must not make the broker blind."""
monkeypatch.delenv(SIGNATURE_ENV, raising=False)
corpse = runtime_dir / "hypr" / "deadbee_1700000001"
corpse.mkdir()
(corpse / "hyprland.log").write_text("this compositor is gone; only its log is here\n")
sock = answering(instance / ".socket.sock")
try:
ipc = HyprlandIPC.discover(runtime_dir)
finally:
sock.close()
assert ipc.request_socket == instance / ".socket.sock"
def test_two_instances_and_neither_answers_refuses(runtime_dir, instance, monkeypatch):
"""Nothing answering is still the dangerous case, and still refused."""
monkeypatch.delenv(SIGNATURE_ENV, raising=False) monkeypatch.delenv(SIGNATURE_ENV, raising=False)
other = runtime_dir / "hypr" / "deadbee_1700000001" other = runtime_dir / "hypr" / "deadbee_1700000001"
other.mkdir() other.mkdir()
bind(instance / ".socket.sock").close() (other / "hyprland.log").write_text("dead\n")
(instance / ".socket.sock").write_text("not a socket")
with pytest.raises(CompositorUnavailable) as caught: with pytest.raises(CompositorUnavailable) as caught:
HyprlandIPC.discover(runtime_dir) HyprlandIPC.discover(runtime_dir)
message = str(caught.value) message = str(caught.value)
assert "2 instances" in message assert "2 instances" in message
assert instance.name in message and other.name in message assert instance.name in message and other.name in message
assert "none of them answered" in message
assert SIGNATURE_ENV in message assert SIGNATURE_ENV in message
def test_two_answering_instances_pick_the_most_recent(runtime_dir, instance, monkeypatch):
"""Two live compositors is a real ambiguity: take the newest, and say so."""
monkeypatch.delenv(SIGNATURE_ENV, raising=False)
older = runtime_dir / "hypr" / "aaaaaaa_1700000001"
older.mkdir()
older_sock = answering(older / ".socket.sock")
newer = runtime_dir / "hypr" / "bbbbbbb_1700000002"
newer.mkdir()
newer_sock = answering(newer / ".socket.sock")
try:
os.utime(older / ".socket.sock", (1_700_000_001, 1_700_000_001))
os.utime(newer / ".socket.sock", (1_700_000_002, 1_700_000_002))
ipc = HyprlandIPC.discover(runtime_dir)
finally:
older_sock.close()
newer_sock.close()
assert ipc.request_socket == newer / ".socket.sock"
def test_two_instances_with_an_explicit_signature(runtime_dir, instance, monkeypatch): def test_two_instances_with_an_explicit_signature(runtime_dir, instance, monkeypatch):
monkeypatch.delenv(SIGNATURE_ENV, raising=False) monkeypatch.delenv(SIGNATURE_ENV, raising=False)
other = runtime_dir / "hypr" / "deadbee_1700000001" other = runtime_dir / "hypr" / "deadbee_1700000001"
+73 -2
View File
@@ -34,6 +34,9 @@ class RecordingSink:
def button(self, output, code, state): def button(self, output, code, state):
self.calls.append(("button", output, code, state)) self.calls.append(("button", output, code, state))
def axis(self, output, axis, value):
self.calls.append(("axis", output, axis, value))
def frame(self, output): def frame(self, output):
self.calls.append(("frame", output)) self.calls.append(("frame", output))
@@ -105,6 +108,51 @@ def test_buttons_follow_the_last_warp_not_the_first(driver, sink):
assert sink.calls[-1] == ("button", "hh-beta", LEFT, DOWN) assert sink.calls[-1] == ("button", "hh-beta", LEFT, DOWN)
# -- axis (scrolling) -------------------------------------------------------
def test_axis_scrolls_the_aimed_output_and_sends_no_implicit_frame(driver, sink):
warp(driver)
sink.calls.clear()
assert driver.handle({"op": "axis", "axis": "vertical", "value": 3}) == {"ok": True}
assert sink.calls == [("axis", "hh-alpha", buttons.AXES["vertical"], 3.0)]
def test_axis_defaults_to_one_vertical_click(driver, sink):
warp(driver)
sink.calls.clear()
assert driver.handle({"op": "axis"}) == {"ok": True}
assert sink.calls == [("axis", "hh-alpha", buttons.AXES["vertical"], 1.0)]
def test_a_horizontal_scroll_is_the_other_axis_and_keeps_the_sign(driver, sink):
warp(driver)
sink.calls.clear()
assert driver.handle({"op": "axis", "axis": "horizontal", "value": -2}) == {"ok": True}
assert sink.calls == [("axis", "hh-alpha", buttons.AXES["horizontal"], -2.0)]
# Two names, two axes -- and positive is down/right, per the protocol rather
# than per us, which is why direction travels in the sign and not the name.
assert buttons.AXES["vertical"] != buttons.AXES["horizontal"]
def test_a_scroll_goes_to_the_pointer_the_last_warp_aimed(driver, sink):
warp(driver, output="hh-alpha")
warp(driver, output="hh-beta")
sink.calls.clear()
driver.handle({"op": "axis", "value": 5})
assert sink.calls == [("axis", "hh-beta", buttons.AXES["vertical"], 5.0)]
def test_scroll_before_any_warp_is_refused(driver, sink):
assert driver.handle({"op": "axis", "value": 3})["ok"] is False
assert sink.calls == [], "a scroll with nothing aimed must emit nothing"
# -- pressed / release_all -------------------------------------------------- # -- pressed / release_all --------------------------------------------------
@@ -220,6 +268,13 @@ def test_a_failed_warp_does_not_change_the_aim():
({"op": "button", "button": "left"}, "'state' must be"), ({"op": "button", "button": "left"}, "'state' must be"),
({"op": "button", "button": "pinky", "state": "down"}, "unknown button 'pinky'"), ({"op": "button", "button": "pinky", "state": "down"}, "unknown button 'pinky'"),
({"op": "button", "button": "left", "state": "sideways"}, "unknown button state"), ({"op": "button", "button": "left", "state": "sideways"}, "unknown button state"),
({"op": "axis", "axis": "diagonal"}, "unknown axis 'diagonal'"),
({"op": "axis", "axis": 0}, "'axis' must be a string"),
({"op": "axis", "value": "a lot"}, "'value' must be a number"),
({"op": "axis", "value": True}, "'value' must be a number"),
({"op": "axis", "value": float("nan")}, "'value' must be finite"),
({"op": "axis", "value": 101}, "wheel-click limit"),
({"op": "axis", "value": -1e9}, "wheel-click limit"),
], ],
) )
def test_malformed_requests_are_answered_not_fatal(sink, request_, fragment): def test_malformed_requests_are_answered_not_fatal(sink, request_, fragment):
@@ -240,5 +295,21 @@ def test_a_non_object_request_is_refused(driver, request_):
assert "must be a JSON object" in driver.handle(request_)["error"] assert "must be a JSON object" in driver.handle(request_)["error"]
def test_the_contracts_five_ops_are_exactly_what_is_implemented(): def test_the_contracts_ops_are_exactly_what_is_implemented() -> None:
assert set(Dispatcher._ops) == {"warp", "button", "frame", "pressed", "release_all"} """The op set cannot drift quietly.
An op that exists here and nowhere in the contract -- or the reverse --
fails this test rather than surprising whoever reads the contract next.
``axis`` is the one addition since the freeze, made deliberately so that a
session can scroll; the set is edited by hand, which is the reminder to
write it down.
"""
assert set(Dispatcher._ops) == {
"warp",
"button",
"axis",
"frame",
"pressed",
"release_all",
}
+57
View File
@@ -213,10 +213,63 @@ class FakeCompositor:
return "ok" return "ok"
if words[:2] == ["output", "remove"] and len(words) == 3: if words[:2] == ["output", "remove"] and len(words) == 3:
return "ok" if self._remove_output(words[2]) else f"no such output {words[2]}" return "ok" if self._remove_output(words[2]) else f"no such output {words[2]}"
if words[:2] == ["dispatch", "movetoworkspacesilent"] and len(words) == 3:
return self._move_to_workspace(words[2])
if words[:2] == ["dispatch", "workspace"] and len(words) == 3:
return self._show_workspace(words[2])
if words[:1] == ["keyword"]: if words[:1] == ["keyword"]:
return self._keyword(command[len("keyword") :].strip()) return self._keyword(command[len("keyword") :].strip())
return f"unknown request {command!r}" return f"unknown request {command!r}"
# -- dispatches -------------------------------------------------------
def _move_to_workspace(self, spec: str) -> str:
"""``<workspace>,address:<addr>`` -- what the broker claims with.
Moving a window is the one thing that puts a real application on its
desk, because no allowlisted app carries the class the rule matches.
"""
workspace_text, _, address = spec.partition(",address:")
if not workspace_text.isdigit() or not address:
return f"malformed move {spec!r}"
client = next((c for c in self._clients if c["address"] == address), None)
if client is None:
return f"no such window {address}"
workspace = int(workspace_text)
monitor = next(
(m["name"] for m in self._monitors if m["id"] == client["monitor"]), None
)
if not any(w["id"] == workspace for w in self._workspaces):
self._workspaces.append(
{"id": workspace, "name": str(workspace), "monitor": monitor, "windows": 0}
)
for entry in self._workspaces:
if entry["id"] == client["workspace"]["id"]:
entry["windows"] = max(0, entry["windows"] - 1)
if entry["id"] == workspace:
entry["windows"] += 1
client["workspace"] = {"id": workspace, "name": str(workspace)}
return "ok"
def _show_workspace(self, workspace_text: str) -> str:
"""``workspace <n>`` -- makes that workspace the visible one.
Only meaningful on the output that holds it, which is how Hyprland
behaves: the workspace's own monitor switches to it.
"""
if not workspace_text.lstrip("-").isdigit():
return f"malformed workspace {workspace_text!r}"
workspace = int(workspace_text)
holder = next((w["monitor"] for w in self._workspaces if w["id"] == workspace), None)
if holder is None:
return f"no such workspace {workspace}"
for monitor in self._monitors:
if monitor["name"] == holder:
monitor["activeWorkspace"] = {"id": workspace, "name": str(workspace)}
return "ok"
def _keyword(self, rest: str) -> str: def _keyword(self, rest: str) -> str:
name, _, body = rest.partition(" ") name, _, body = rest.partition(" ")
if name == "workspace": if name == "workspace":
@@ -353,6 +406,10 @@ class RecordingSink:
self._check() self._check()
self.calls.append(("button", output, code, state)) self.calls.append(("button", output, code, state))
def axis(self, output: str, axis: int, value: float) -> None:
self._check()
self.calls.append(("axis", output, axis, value))
def frame(self, output: str) -> None: def frame(self, output: str) -> None:
self._check() self._check()
self.calls.append(("frame", output)) self.calls.append(("frame", output))
@@ -276,6 +276,47 @@ def test_container_reconcile_reports_the_difference_it_repaired(rig: Rig):
assert rig.daemon.session(session_id)["window"] is None assert rig.daemon.session(session_id)["window"] is None
def test_a_window_that_lands_outside_its_desk_is_moved_onto_it(rig: Rig):
"""The case the class-matching rule could never handle.
Every allowlisted application keeps its own class, so the per-session rule
matches nothing and the window is swept to the quarantine backstop — a desk
created and left empty. The compositor reports each client's pid, so the
window can still be claimed: moved onto its own desk's workspace, with that
workspace made the visible one, and read back afterwards to say so.
"""
rig.compositor.auto_window = False
session_id = start(rig.daemon, "sleeper")
# It turns up on the main screen instead — where a quarantined window goes.
rig.compositor.add_window(f"hh-{session_id}", 1, "HEADLESS-1", pid=WINDOW_PID)
rig.daemon.call("container.reconcile", id=session_id)
record = rig.daemon.session(session_id)
desk = record["container"]
assert record["workspace"] == desk["workspace"], "the read-back says where it is"
assert record["foreground"] is True, "its desk is showing it"
assert desk["state"] != "degraded"
commands = rig.plan_commands()
assert (
f"dispatch movetoworkspacesilent {desk['workspace']},"
f"address:{record['window']['address']}" in commands
), commands
assert f"dispatch workspace {desk['workspace']}" in commands, commands
def test_a_window_already_on_its_desk_is_not_moved(rig: Rig):
"""The steady state costs nothing: a move is a real change on screen."""
session_id = start(rig.daemon, "sleeper")
rig.daemon.call("container.reconcile", id=session_id)
assert rig.daemon.session(session_id)["workspace"] == 1000
assert not [c for c in rig.plan_commands() if c.startswith("dispatch")]
# =========================================================================== # ===========================================================================
# the start-up scan # the start-up scan
# =========================================================================== # ===========================================================================
@@ -28,6 +28,26 @@ def compositor(short_tmpdir: Path):
fake.stop() fake.stop()
def test_a_disabled_host_is_unavailable_and_says_why() -> None:
"""Turning containers off must be configuration, not a false compositor.
``available()`` is the whole mechanism: the manager already treats an
unavailable host as "sessions without containers", which is exactly what
turning them off means, so nothing else has to branch on it. It must also
not ask a compositor this deployment never intended to talk to -- and its
reason has to name the key, or a machine with containers disabled looks
identical to one whose compositor is down and the operator goes hunting for a
fault that is not there.
"""
host = ContainerHost(
None, allocator=containers.WorkspaceAllocator(1000, 4), enabled=False
)
assert host.available() is False
assert "containers_enabled" in (host.last_error or "")
def make_host(fake: FakeCompositor, *, slots: int = 4) -> ContainerHost: def make_host(fake: FakeCompositor, *, slots: int = 4) -> ContainerHost:
return ContainerHost( return ContainerHost(
fake.runtime_dir, fake.runtime_dir,
+104
View File
@@ -24,6 +24,10 @@ from hhstreamd.registry import ContainerState, Session, State, container_record
LEFT = 0x110 LEFT = 0x110
DOWN, UP = 1, 0 DOWN, UP = 1, 0
#: ``wl_pointer_axis``, mirrored from the driver for the same reason the button
#: codes above are: these tests assert the events, so they name the wire values.
VERTICAL, HORIZONTAL = 0, 1
def make_session(session_id: str = "foot-aaaaaa", *, container: bool = True) -> Session: def make_session(session_id: str = "foot-aaaaaa", *, container: bool = True) -> Session:
return Session( return Session(
@@ -590,3 +594,103 @@ def test_the_sweeper_expires_a_lease_nobody_came_back_for(short_tmpdir: Path):
if manager.driver is not None: if manager.driver is not None:
manager.driver.close() manager.driver.close()
fake.stop() fake.stop()
# ---------------------------------------------------------------------------
# scrolling
# ---------------------------------------------------------------------------
def test_scroll_drives_axis_then_frame_on_the_sessions_own_output(harness):
harness.add()
lease = harness.run(harness.manager.pointer_acquire("foot-aaaaaa"))["lease_id"]
harness.fake.sink.calls.clear()
reply = harness.run(harness.manager.pointer_scroll("foot-aaaaaa", lease, "vertical", 3))
assert (reply["axis"], reply["amount"]) == ("vertical", 3.0)
assert harness.fake.sink.calls == [
("axis", "hh-foot-aaaaaa", VERTICAL, 3.0),
("frame", "hh-foot-aaaaaa"),
]
def test_scrolling_up_is_the_sign_and_not_a_different_verb(harness):
"""One verb, one axis name, and the protocol's sign carries the direction."""
harness.add()
lease = harness.run(harness.manager.pointer_acquire("foot-aaaaaa"))["lease_id"]
harness.fake.sink.calls.clear()
harness.run(harness.manager.pointer_scroll("foot-aaaaaa", lease, "vertical", -3))
assert harness.fake.sink.calls == [
("axis", "hh-foot-aaaaaa", VERTICAL, -3.0),
("frame", "hh-foot-aaaaaa"),
]
def test_a_horizontal_scroll_uses_the_other_axis(harness):
harness.add()
lease = harness.run(harness.manager.pointer_acquire("foot-aaaaaa"))["lease_id"]
harness.fake.sink.calls.clear()
harness.run(harness.manager.pointer_scroll("foot-aaaaaa", lease, "horizontal", 2))
assert harness.fake.sink.calls == [
("axis", "hh-foot-aaaaaa", HORIZONTAL, 2.0),
("frame", "hh-foot-aaaaaa"),
]
def test_a_scroll_restamps_the_resume_point_without_moving_it(harness):
"""The cursor stays put -- and the record must not pretend otherwise."""
session = harness.add()
lease = harness.run(harness.manager.pointer_acquire("foot-aaaaaa"))["lease_id"]
before = dict(session.pointer["bookmark"])
harness.clock.advance(7.0)
reply = harness.run(harness.manager.pointer_scroll("foot-aaaaaa", lease, "vertical", 3))
assert (reply["bookmark"]["x"], reply["bookmark"]["y"]) == (before["x"], before["y"])
assert reply["bookmark"]["output"] == before["output"]
assert reply["bookmark"]["at"] != before["at"], "the session acted, so the stamp moved"
assert harness.manager.pointer.state()["holder"] == "foot-aaaaaa"
@pytest.mark.parametrize("bad", [0, 0.0, 101, -101, "three", True, None])
def test_a_scroll_amount_off_the_wire_is_bounded(harness, bad):
harness.add()
lease = harness.run(harness.manager.pointer_acquire("foot-aaaaaa"))["lease_id"]
with pytest.raises(ManagerError) as exc:
harness.run(harness.manager.pointer_scroll("foot-aaaaaa", lease, "vertical", bad))
assert exc.value.error == "invalid_argument"
def test_an_unknown_axis_is_refused_by_the_driver_not_invented(harness):
harness.add()
lease = harness.run(harness.manager.pointer_acquire("foot-aaaaaa"))["lease_id"]
harness.fake.sink.calls.clear()
with pytest.raises(ManagerError) as exc:
harness.run(harness.manager.pointer_scroll("foot-aaaaaa", lease, "diagonal", 3))
assert exc.value.error == "driver_refused"
assert harness.fake.sink.calls == [], "a refused scroll must emit nothing"
def test_a_scroll_without_the_lease_is_refused(harness):
session = harness.add()
harness.run(harness.manager.pointer_acquire("foot-aaaaaa"))
harness.fake.sink.calls.clear()
with pytest.raises(ManagerError):
harness.run(
harness.manager.pointer_scroll("foot-aaaaaa", "not-this-lease", "vertical", 3)
)
assert harness.fake.sink.calls == []
assert session.pointer.get("bookmark") is not None, "a refused scroll changes nothing"
+78
View File
@@ -0,0 +1,78 @@
"""The config file's non-allowlist half.
These keys are rendered by the NixOS module and, until this existed, read by
nothing: a deployment could turn containers off or move a lease deadline and be
silently ignored, while the defaults happened to make it look honoured. The
values here are therefore the ones the module actually writes -- if the module's
shape changes, these tests are where that shows up.
"""
from __future__ import annotations
from pathlib import Path
import tomllib
import pytest
from hhstreamd.policy import PolicyError, load_settings, settings_from_dict
#: Exactly what `nix/module.nix` renders today, through
#: `services.hyprland-headless-stream`, with containers turned off.
MODULE_SHAPE = """
[defaults]
max_sessions = 4
idle_timeout_s = 10800
max_containers = 2
[compositor]
runtime_dir = "/run/hypr-stream"
containers_enabled = false
workspace_base = 1000
container_mode = "1280x720"
[pointer]
enabled = true
lease_ttl_s = 30
lease_max_s = 300
[apps.chromium]
exec = "/nix/store/x/bin/chromium"
args_allow = ["--ozone-platform=wayland"]
"""
def test_the_module_shape_is_read() -> None:
settings = settings_from_dict(tomllib.loads(MODULE_SHAPE))
assert settings.containers_enabled is False
# max_containers lives in [defaults] while the rest live in [compositor];
# that split is the module's, and reading only one of them is how a budget
# silently stops applying.
assert settings.container_slots == 2
assert settings.container_base == 1000
assert settings.container_mode == (1280, 720)
assert (settings.pointer_ttl_s, settings.pointer_max_s) == (30.0, 300.0)
def test_containers_default_to_on_when_the_key_is_absent() -> None:
"""A config written before this setting existed keeps working."""
assert settings_from_dict({"apps": {}}).containers_enabled is True
def test_a_bad_mode_is_refused_by_name() -> None:
"""A typo must be loud: a container at the wrong size looks like an app bug."""
with pytest.raises(PolicyError) as exc:
settings_from_dict({"compositor": {"container_mode": "wide"}})
assert "container_mode" in str(exc.value)
def test_an_absent_file_is_the_defaults(tmp_path: Path) -> None:
"""Missing config behaves conservatively rather than refusing to start."""
settings = load_settings(tmp_path / "absent.toml")
assert settings.containers_enabled is True
assert settings.container_mode == (1280, 720)
+1 -1
View File
@@ -56,7 +56,7 @@ def test_server_defaults_match_the_contract():
args = server.build_parser().parse_args([]) args = server.build_parser().parse_args([])
assert args.socket == "/run/hypr-stream/stream.sock" assert args.socket == "/run/hypr-stream/stream.sock"
assert (args.fps, args.format, args.capture_command) == (30, "png", None) assert (args.fps, args.format, args.capture_command) == (30, "png", None)
assert server.DEFAULT_CAPTURE_COMMAND == ["grim", "-t", "png", "-"] assert server.DEFAULT_CAPTURE_COMMAND == ["grim", "-c", "-t", "png", "-"]
def test_capture_command_is_one_argv_token_per_flag(): def test_capture_command_is_one_argv_token_per_flag():