Kernel: a kill never waits on its victim — the last thread out tears its process down - #549

Merged

Japabu merged 30 commits into

mainfrom

wt/toyos-killfix

Sep 28, 2026

Commits on Sep 27, 2026

kernel: a kill posts its victim's retires and returns; the reaper finishes it

Japabu
and
claude
committed
8864321
proclife: KillEachOther gives each process a second thread, and a sibling is a killer

Japabu
and
claude
committed
b88bf98
tests: quiesce_stops_the_machine counts the reaper among the threads the stop names

Japabu
and
claude
committed
db57dea
Merge origin/main
Japabu
committed
61c0a7f
kernel: the reaper takes whichever kill is released, and the teardown asserts it

Japabu
and
claude
committed
1ae6558
Merge origin/main

Japabu
and
claude
committed
a7457a3
Merge remote-tracking branch 'origin/main' into wt/toyos-killfix
Japabu
committed
c88bb4a
review #549 round 2: no kernel thread's pid opens, and the reaper's choice is the model's

Japabu
and
claude
committed
1c5ca17
Merge remote-tracking branch 'origin/main' into wt/toyos-killfix
Japabu
committed
b4d7ef6
kernel: the last thread out tears its process down; no reaper, no retire wait

Japabu
and
claude
committed
fa1e325

Commits on Sep 28, 2026

Merge origin/main into wt/toyos-killfix

Japabu
and
claude
committed
456d8cb
kernel: the last thread out is in its process until its teardown is done

Japabu
and
claude
committed
a6aaf3e
kernel: a process's end is not pollable

Japabu
and
claude
committed
57c89b5
kill_while_blocked: arm 4 waits for its spinner's exit, with no clock

Japabu
and
claude
committed
179f41e
kernel-loom: the nested serve is a thread of its own

Japabu
and
claude
committed
0c48340
Merge origin/main into wt/toyos-killfix

Japabu
and
claude
committed
393a4f2
quiesce-last: the hold ends on a sweep that counted the held thread alone

Japabu
and
claude
committed
63d0b08
scheduler: a killed thread's teardown at the Ring 3 boundary runs with IF set

Japabu
and
claude
committed
a592498
kill_ends_every_wait: a ceiling red names the arm; the arm race is filed

Japabu
and
claude
committed
751e36d
A refused hold releases its slot, a gave-up stop names its threads, and sleep leads the wait arms

Japabu
and
claude
committed
009fa2a
Merge origin/main into wt/toyos-killfix

Japabu
and
claude
committed
ce95b55
quiesce_wakes_on_the_last_teardown shares quiesce_wakes_on_the_last_park's disabled issue

Japabu
and
claude
committed
4d612d6
kill_ends_every_wait kills a child only once the roster shows it parked

Japabu
and
claude
committed
0f28ea0
Merge origin/main into wt/toyos-killfix

Japabu
and
claude
committed
ae06655
kill_ends_every_wait's roster wait has no clock of its own

Japabu
and
claude
committed
5163e9b
Round 8 review fixes: the roster poll fails by name, not by the harness ceiling, on a header over 256 entries or a main thread gone/zombied; file the missing roster-entry decoder as its own issue; …

Japabu
and
claude
committed
82f917b
Merge remote-tracking branch 'origin/main' into wt/toyos-killfix
Japabu
committed
80276cb
Merge origin/main (#564, the measured schedule) into wt/toyos-killfix

Japabu
and
claude
committed
4e67c19
Round of #549 review: delete four stale lists the merge rewrote instead of removing

Japabu
and
claude
committed
4c2e052
Merge remote-tracking branch 'origin/main' into wt/toyos-killfix

Japabu
and
claude
committed
254f1fe

Navigation Menu

Kernel: a kill never waits on its victim — the last thread out tears its process down - #549

Kernel: a kill never waits on its victim — the last thread out tears its process down#549
Japabu merged 30 commits into
mainfrom
wt/toyos-killfix

Commits on Sep 27, 2026

kernel: a kill posts its victim's retires and returns; the reaper finishes it

proclife: KillEachOther gives each process a second thread, and a sibling is a killer

tests: quiesce_stops_the_machine counts the reaper among the threads the stop names

Merge origin/main

kernel: the reaper takes whichever kill is released, and the teardown asserts it

Merge origin/main

Merge remote-tracking branch 'origin/main' into wt/toyos-killfix

review #549 round 2: no kernel thread's pid opens, and the reaper's choice is the model's

Merge remote-tracking branch 'origin/main' into wt/toyos-killfix

kernel: the last thread out tears its process down; no reaper, no retire wait

Commits on Sep 28, 2026

Merge origin/main into wt/toyos-killfix

kernel: the last thread out is in its process until its teardown is done

kernel: a process's end is not pollable

kill_while_blocked: arm 4 waits for its spinner's exit, with no clock

kernel-loom: the nested serve is a thread of its own

Merge origin/main into wt/toyos-killfix

quiesce-last: the hold ends on a sweep that counted the held thread alone

scheduler: a killed thread's teardown at the Ring 3 boundary runs with IF set

kill_ends_every_wait: a ceiling red names the arm; the arm race is filed

A refused hold releases its slot, a gave-up stop names its threads, and sleep leads the wait arms

Merge origin/main into wt/toyos-killfix

quiesce_wakes_on_the_last_teardown shares quiesce_wakes_on_the_last_park's disabled issue

kill_ends_every_wait kills a child only once the roster shows it parked

Merge origin/main into wt/toyos-killfix

kill_ends_every_wait's roster wait has no clock of its own

Round 8 review fixes: the roster poll fails by name, not by the harness ceiling, on a header over 256 entries or a main thread gone/zombied; file the missing roster-entry decoder as its own issue; …

Merge remote-tracking branch 'origin/main' into wt/toyos-killfix

Merge origin/main (#564, the measured schedule) into wt/toyos-killfix

Round of #549 review: delete four stale lists the merge rewrote instead of removing

Merge remote-tracking branch 'origin/main' into wt/toyos-killfix