Every tab is maximized and kick draws no frames, so new toplevels
are pinned to client-side mode (headerbars keep working) and later
server-side requests are ignored. Pinning happens at map -- set_mode
asserts on a never-committed surface, and decoration-before-commit is
the normal order (calling it in new_toplevel_decoration aborts
wlroots); late binds pin immediately when already initialized.
Adds an xdeco probe (maps one shm-backed toplevel, reports the
configured mode) with a regression test asserting mode=1, tracks it
in build.sh staleness, and covers the global in protocol_globals.
Full suite: 153 passed, 2 skipped.
All 16 assert _wait_log(s, rx, timeout) sites become s.wait_for(rx,
timeout); every caller only asserted truthiness, and both poll the
whole log identically, so behavior is unchanged (failures now report
the pattern plus a log tail instead of a custom one-liner). Sites on
_wait_log's 15s default keep timeout=15.0 explicitly since wait_for
defaults to 20s. The helper is deleted. Full suite: 152 passed,
2 skipped.
146 assert X.proc.poll() is None spellings (bare, log-tail tails of
varying length, custom operation messages) become _assert_alive(s,
msg): custom messages are preserved up front, the log tail is always
attached. The 4 crash-indicator loops become _assert_no_crash(log)
over the union tuple. Full suite: 152 passed, 2 skipped.
Mechanical whitespace-only follow-up to phases 1-2: continuation
lines that were paren-aligned to the old s = _spawn_* heads now align
to the with _run_* heads. Verified with git diff -w (empty).
28 deadline/while/exists polling loops become one-line
assert _wait_out(path, contains, timeout) calls; the helper retries
empty/partial reads and survives exists/read races, and callers keep
their own assert messages (including the live '(missing)' payload
diagnostics). Deliberately untouched: in-loop liveness asserts
(fail-fast under load), pixel/dump/log-touch loops, line-count gates,
XWayland DISPLAY export loops, and the _oneshot_*/_poll_* helpers.
Full suite: 152 passed, 2 skipped.
101 single-session sites go from s = _spawn_* + try/finally
s.close() to with _run_*() as s:. Standard boot preambles fold into
wait= (any/any_frame/app0/app0_mapped). Client-cleanup tails are kept
as inner try/finally so kill order is unchanged; the dead
if poll() is None: terminate() after close() is dropped (close already
terminate+wait/kills). Fixtures, kick-in-kick outer/inner pairs and
_run_* internals are untouched. Full suite: 152 passed, 2 skipped.
Deeper-X11 instrumentation (permanent, DEBUG): every associate event
now logs surface null/mapped/hooked state, classify logs wid/xpid/pgid
at entry, and pending-map has its own line -- closing the blind spot
between xwm pairing and tab claim. Plus kick_monotonic_ms() timestamps
on the new lines for stall measurement (prefix-safe for all test
regexes).
Paid off immediately: two stalls autopsied to paired-but-never-mapped
wlr_surfaces (associate ran with surface mapped=0, no map for 15-30s,
classify never invoked -- proven by the absent entry line). The stall
is therefore upstream of kick classification (X MapNotify delivery /
xwm mapping under -n15), not in it; locating it further needs live
capture during a stall, not post-hoc logs.
test_xwayland_ready_and_claim failed once the same way (paired xmessage
surface never mapped within 15s): marked flaky per the seen-it rule
(passes serially).
The 15s os.kill poll asked for trouble (fixed window on a foreign event
loop). Replaced with kick's own SIGCHLD reap line ([child] pid N
exited, written unbuffered per reaped child) -- the same signal, event
anchored, 30s backstop only.
Two bugs found along the way, both in the test not the product:
- claim pid != reaped pid (sh -c forks; the claim is the grandchild,
the reap is the direct-child launcher) -- wait on [launch] app=0's pid.
- [child] uses 'pid ' (space) vs [launch]/[claim] 'pid=' -- the first
version of this wait matched nothing for 30s, deterministically.
(Also: earlier validation runs silently deselected this marked test;
always pass -m flaky when validating marked tests.)
Unmarked after 12/12 at -n15 + serial green.
hold_alt_down_repeats + alt_tab_hold_repeats asserted repeat COUNTS in
fixed windows -- pure wall-clock. Now: press -> wait [repeat] start
(new DEBUG lifecycle lines in kbd_repeat_start/cancel; stop logged only
when disarming a live timer) -> wait for ONE synthesized repeat
(count-based, unbounded time) -> release -> wait [repeat] stop ->
immediate count assert (no timer exists post-disarm, so no silence
window needed). Generous timeouts bound hangs only; slow boxes take
longer without failing a rate. hold_alt unmarked (6/6 at -n15).
xwayland_alt_delete_closes_x11_tab: 1 failure in 13 runs, no body --
marked flaky per the seen-it rule.
Evidence: -n15 gate clean (153 passed), serial gate clean (153).
The 7 host-timed tests failed not (only) on timeouts: autopsy showed
parallel workers stealing each other's live-host clipboard (wl-paste
returns another test's token -- no timeout survives that).
Clipboard x3: the failure is cross-talk on ONE shared host clipboard.
Serialize just those sections with an flock file across xdist workers
(the 45s polls stay as backstops for genuine host slowness -- marks
kept; focus-based backend still needs the real host).
Fullscreen x3 + cursor-shape: migrate to per-test outer-kick hosts
(same xdg/fullscreen/shape protocols, deterministic compositor instead
of GNOME). No coverage lost (single code path; timing was the only
host variable). cursor-shape uses deterministic warp-enter; fullscreen
uses tap/chord acks. Added KICK_REPEAT_DELAY_MS + extra_args to
_spawn_nested_in. Unmarked after -n15 validation.
hold_alt stays (wall-clock timer).
Evidence: -n15 gate 3x clean (152 passed), serial gate clean (152),
migrated tests 6/6 at -n15. 12 -> 7 flaky.
Killing the embedded Xwayland server aborted the compositor:
surface_handle_resource_destroy assert (non-empty map listener list)
via wl_client_destroy in the main loop's flush_clients.
Mechanism: the dead server's Wayland client connection is purged before
wlroots' xwm cascade dissociates anything, so wlr_surfaces die with our
map/unmap/commit listeners still attached.
Fix: watch the Xwayland Wayland client (its destroy signal fires before
its resources are purged) and proactively detach surface listeners only
-- across claimed windows, override-redirect windows and unclaimed
pendings -- leaving xsurface-level listeners for the later tab teardown.
Idempotent; later dissociate/xsurface-destroy find clean flags. The
restart test now also waits for the X11 window's map before killing, so
the crash path (listeners attached at death) is deterministically
exercised instead of raced.
Also re-mark output_mgmt_apply flaky (1 post-fix assertion under -n15;
budgets help but the modeset roundtrip stays load-sensitive).
Evidence: restart test 12/12 at the -n15 repro config (was ~2/9
crashing), detach handler firing on every server death, -n15 gate
clean, serial gate clean (148), flaky set green (11).
Fixed + unmarked (all green serially, holding at -n15):
- dead_command: wait for the exit line (event) before proc.wait instead
of a fixed 10s reap timeout.
- output_mgmt_apply: phase on the existing [output] apply line so the
commit wait covers only client reconfigure lag (30s budgets).
- gestures_roundtrip: cursor-line + [gesture] forward-line acks per
sequence (offset), single client-file poll; no more resend-on-slow.
- x11_client_requests_driven: phased [xwayland] ready wait (own budget)
+ startup-crown settle/nav.
Hardened, marks kept (host-timed roundtrips): 3 fullscreen tests use
tap_ack/chord_ack + 30s host-configure waits.
Marked flaky (post-fix sightings under load):
- x11_menu_popup (associate never lands in X11-heavy -n15 runs),
- x11_leader (1 claim timeout post-fix),
- xwayland_server_restart_survived (real SIGSEGV in
wl_display_flush_clients on Xwayland SIGKILL restart -- pre-existing
race, unrelated to this work; needs a proper fix, not a test tweak).
Net flaky 14 -> 11. Evidence: -n15 gate 2x clean (148 passed),
serial gate clean, flaky set green serially.
test_x11_menu_popup_attaches_as_member failed 3x under parallel load
with no diagnosed mechanism: marked flaky per the seen-it-fail rule
(10 flaky total; all pass serially).
Sleep follow-up: fixed sleeps 131 -> 112 (-26 assertion-guarding,
+poll intervals). Converted the safe remainder with the same
techniques -- cursor-line waits for pointer baselines, tap_ack quorums
for absence guards (function keys, plain keys, minimize/move/resize),
client-file polling for the maximize-refusal roundtrip, dropped the
now-vacuous reorder sleep (consumed chords + stretched repeats leave no
async transition source). Left deliberately: wall-clock timers (idle,
hold-settle), teardown-survival settles in stable tests, X11 structural
sleeps, rate-limit pokes, poll intervals.
Verified: -n15 gate clean 4x (149 passed), -n15 full run fails only the
10 pre-marked flaky, flaky set green serially.
Mechanisms found by autopsying failing sessions' logs, each fixed with
logging + event-anchored waits instead of sleeps/fixed assumptions:
- Input-consumed acks: new Session.tap_ack/chord_ack snapshot the log
offset BEFORE sending (send-then-snapshot races a descheduled test and
misses its own line) and wait for the existing [kbd] press line.
wait_for_new_after covers release-side effects ([mru] commit).
- Wall-clock repeats displacing exact MRU/cycle/reorder state: a tap whose
press/release processing spans the 600ms repeat delay catches synthesized
repeats (autopsied: 3 preview steps from 1 press). KICK_REPEAT_DELAY_MS
env override (default 600); all test sessions stretch to 60s except the
two hold-to-repeat tests, which get private sessions with real repeats.
- Startup-crown flips (last mapper promotes itself): all crown reads
settled (both maps = quiescence) + relative navigation; black-tab crowns
log [diag] active slot= (no active app= line) -- waited accordingly.
- Title/data decoupling: [title] now emitted at data sources (map-end with
final title, post-refocus destroy, renames, focus promotion, subcycle,
tag-group join), not just overlay draws; premature claim-time hook tried
and reverted (wrong pick poisons the dedup cache).
- Pointer tests: new [cursor] pos= DEBUG line on warp/motion; dump-poll
loops become cursor-line + single-dump asserts. Also fixed a latent
_spawn_standalone bug: env -i dropped KICK_DUMP_DIR, so parallel
standalone sessions shared /tmp/host_N.ppm and read each other's dumps.
- Spawn races: _nested_session/no_args retry instant client death.
- X11/layer/modal: settle+nav, map-anchored claims, content dump-polls,
settle-poll instead of fixed sleeps.
Genuine product race fixed: fork->setpgid window let SIGCHLD-driven
sweep_dead_startup_groups conclude ESRCH on a not-yet-existing group and
drop a LIVE tab (orphaned sleep survivors found on the box). Parent-side
setpgid + pid-liveness double-check before sweeping.
Remaining 9 flaky (wall-clock timers, host-compositor roundtrips):
gestures, hold_alt, fullscreen x3, clipboard host x3, cursor-shape.
Evidence: -n10 clean 5x (150 passed, ~28s), serial clean (150 passed),
flaky set green (9 passed).
Pilot A (input-consumed signal): test_close_black_tab_active_last waits
for [kbd] key=111 state=1 after the chord, proving the compositor
consumed the key before waiting for the effect. Unmarked flaky.
Pilot B (data-source logging): [title] was only emitted from overlay
draw passes, so title tests needed Alt held + live frames. tab_row_title
is now also called at the data-source points where the topmost window's
title can change -- end of the map handler (first commit carries the
final title with tab assigned and focus settled), after refocus in
window_destroy_common, on set_title renames, and on the focus-promotion
path -- so row titles are log-observable with no overlay frames.
test_sub_windows_stack_mode_row_title_follows_top drops the Alt hold and
is unmarked. (A premature hook at claim time was tried and reverted: the
title has not arrived yet there, and resolving with the wrong pick
poisons that window's dedup cache and kills the later fallback line.)
Startup-crown races (found when full-suite validation failed 9-19 tests
that pass in isolation): each tab's first map promotes it, so under load
the LAST mapper steals the crown and neither the transient
wait_any_active read nor 'settled crown == 1' holds. Five tests now
settle (both tabs mapped = quiescence: no more first-maps, no more
promotions) and navigate relatively: alt_tab_walks/commit/1/repeats and
the sigkill-trap test.
-n10 clean 3x (123 passed, ~21s); serial clean (123 passed).
./test.sh with no -n/-m runs the suite as: (1) non-flaky tests under
-n10 (~25s), (2) flaky-marked tests serially (~35s); either phase
failing fails the run (~60s total vs ~160s serial before). Explicit
-n/-m flags keep the old single-phase passthrough behavior
(./test.sh -n0 forces serial). The -m override in phase 2 restates
'not fuzz' since command-line -m replaces pytest.ini's -m.
Parallel runs (-n10, needed since -n20 overwhelms the desktop
compositor) exposed load-sensitive tests as failures. Fix the
fixable, mark the rest flaky (all marked tests pass serially;
./test.sh -m flaky stays green in ~33s):
Systemic fixes (free on success -- poll loops exit early):
- Session poll-loop timeouts retuned for loaded boxes: wait_for
15->20s, wait_any_active/active_frame 10->20s, wait_app_active/
wait_surface_mapped 10->15s, wait_active/change 3->10s,
cycle_ack 5->15s; drop explicit 3s waits onto the new defaults.
- Startup-crown race: wait_any_active could catch a transient crown
while the other tab was still pending, so four tests now settle
both maps before reading the crown; alt_shift_t quiesces maps
before snapshotting the transition count.
- Pointer-constraint and output-mgmt poll deadlines 5/20->15/30s
(per-iteration dump+parse cost under load, not logic).
Flaky marks (all pass serially; fail only via load timing -- key
chords/taps landing in shifted state, overlay/pixel timing, host
roundtrips over the shared desktop compositor, X11 startup races,
repeat-rate wall clocks, multi-client dances):
hold_alt, arrows_reorder, sub_windows x4, fullscreen x3,
clipboard x3, pointer x3, screen_mode, layer x2, x11 x3,
startup-crown, modal, quit, plain_delete, cycle_wraps,
function_keys, alt_delete x2, tab_death, gestures_roundtrip,
modifier_super, no_args, alt_shift_move, close_black_tab,
output_mgmt_apply, screen_mode_sets_output_size.
Make ./test.sh -n4 work (~47s vs ~156s serial):
- kick.c: KICK_DUMP_DIR env override for frame-dump .ppm paths
(default /tmp); per-boot dump_seq counters no longer collide
across concurrent compositors.
- tests: every spawn gets a unique log file and a per-session dump
directory owned by the Session (close removes the tree instead of
globbing /tmp, which raced parallel workers' live dumps).
- 14 protocol-globals tests share one module-scoped keeper session
via short-lived --globals-out clients instead of 14 boots.
- Two presence-waits converted from fixed sleeps to event waits.
- Harden the fullscreen round-trip: wait for seat keyboard focus
([diag] kbd-enter, new debug line in window_enter) before tapping
F -- map alone does not guarantee enter under load and the tap was
silently dropped.
Serial remains the deterministic gate (documented in test.sh and
README): a few pre-existing timing-sensitive tests can still flake
under parallel load; they pass alone and serially.
Stock wl-clipboard only speaks the legacy zwlr_data_control protocol
kick no longer offers (removed in wlroots 0.20), and its wl_data_device
fallback needs a focus serial a surfaceless tool never receives, so the
six clipboard tests poll-timed-out on hosts where kick is the data-control
endpoint (e.g. labwc).
wltest gains --clipboard-set/--clipboard-demand-expect,
--dc-primary-set/--dc-primary-demand-expect and --dc-clear over
ext-data-control-v1 (no focus or serial needed; setters exit 0 on
cancel like wl-copy -f), and the tests use them on kick sockets while
wl-copy/wl-paste stay on the real-host side.
Booting a fresh kick per globals assertion wastes a spawn each for
zero benefit: the advertised global set is static and these tests read
only the --globals-out file. A module-scoped globals_session fixture
boots one keeper compositor; each of the 11 globals tests runs a
short-lived --globals-out client against its socket instead
(11 spawns -> 1).
Groundwork for future parallel runs: every spawn now gets a unique
log file (_fresh_log) instead of sharing /tmp/tabtest.log, so workers
can no longer unlink or interleave each other's logs.
Also convert two presence-waits from fixed sleeps to event waits
(x11 popup attach waits for the associate line; maximize-refusal waits
for the first configure line): faster and immune to latency under
load. Crash-guard and timeout-crossing sleeps are deliberately
untouched.
Three of the four protocols new in wlroots 0.20 (ext-workspace has no
meaningful mapping onto kick's tab model, so it stays out):
- wl_fixes (core, v1): advertise the global; wlroots serves
destroy_registry itself. Clients like Firefox use it to tear down
registries without tripping protocol errors.
- xdg-toplevel-tag-v1: record each window's tag + translated
description ([tag] log line per change), show 'title [tag]' in the
overlay, and with --tab-groups on stack an unmapped external window
onto the tab whose window carries the same tag (browser-style tag
groups). Tags arriving before claim wait in a pending list keyed by
toplevel; mapped windows never move.
- color-representation-v1: advertise an explicit SDR sRGB set
(premultiplied-electrical alpha, identity coefficients in full
range) rather than the renderer's query, which reports empty on
GLES2; wlroots validates client values against it.
Tests: wltest gains --tag-set/--tag-desc and --color-probe/--color-out
(bind-time listener, since the manager bursts on bind); the globals
test covers all three new globals; new tag, tag-grouping and color
round-trip tests; README documents all three.
The clipboard bridge backend has been ext-data-control since the
wlr-data-control migration; three comments/log strings still named the
old protocol. Cosmetic-only, no behavior change.
The wlr-screencopy unstable protocol is deprecated; its superseding
staging protocols ext-image-copy-capture-v1 and
ext-image-capture-source-v1 ship in wayland-protocols. Migrate:
- advertise wlr_ext_image_copy_capture_manager_v1 and
wlr_ext_output_image_capture_source_manager_v1 instead of the
wlr_screencopy_manager_v1 global (wlroots runs the whole session/frame
flow off the output commit path, so no listener wiring is needed)
- rewrite wltest --screencopy for the new client flow: bind the capture
managers, create an output capture source, open a copy-capture session,
wait for the constraint batch (buffer_size/shm_format/done), then attach
a shm buffer and ask for one frame
- generate the two client stubs from wayland-protocols and drop the
bundled wlr-screencopy-unstable-v1.xml
- update the globals test to the new interface names
Existing wl_data_device-based clipboard bridge code is unaffected.
The wlr-data-control unstable protocol is deprecated; its superseding
staging protocol ext-data-control-v1 ships in wayland-protocols.
Migrate the host-side clipboard/primary bridge to it:
- bind ext_data_control_manager_v1 version 1 instead of zwlr_manager
(ext-data-control v1 always supports primary selection, so the
dc_primary_supported gate is dropped and the registry bind is plain)
- rename all zwlr_data_control_* to ext_data_control_* types/calls
- generate ext-data-control-v1 from wayland-protocols and drop the
bundled wlr-data-control-unstable-v1.xml
- update README and test log-message expectations
wltest --screencopy is unaffected; the screencopy migration is separate.
