Skip to content

Fix Linux signal stderr cleanup #208

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
yihanzhu merged 2 commits into from
Sep 1, 2026
Merged

Fix Linux signal stderr cleanup #208

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cfd0218
Avoid Linux Bash trap-table race
Aug 31, 2026
50b24b5
Harden group-wide signal cleanup
Aug 31, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
84 changes: 61 additions & 23 deletions adapter-tests/v1/runner.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,12 +30,21 @@ fixture_root=$(CDPATH='' cd -P -- "$fixture_root" && pwd -P) || runner_error E_F

run_tmp=$(/usr/bin/mktemp -d "$fixture_root/scratch/run.XXXXXX") || runner_error
cleanup() {
local deadline
if [ -n "${run_tmp:-}" ]; then
/bin/rm -rf -- "$run_tmp"
deadline=$((SECONDS + 3))
while { [ -e "$run_tmp" ] || [ -L "$run_tmp" ]; } && [ "$SECONDS" -lt "$deadline" ]; do
/bin/rm -rf -- "$run_tmp" 2>/dev/null || :
[ ! -e "$run_tmp" ] && [ ! -L "$run_tmp" ] && return 0
/bin/sleep 0.01 || :
done
/bin/rm -rf -- "$run_tmp" 2>/dev/null || return 1
[ ! -e "$run_tmp" ] && [ ! -L "$run_tmp" ]
fi
}
ACTIVE_CHILD_GROUP=''
ACTIVE_CHILD_PID=''
CHILD_SEQUENCE=0
SIGNAL_DEFER=0
SIGNAL_EXITING=0
PENDING_SIGNAL_STATUS=0
Expand All @@ -45,30 +54,41 @@ group_alive() {
terminate_active_group() {
local group=${ACTIVE_CHILD_GROUP:-}
local leader=${ACTIVE_CHILD_PID:-}
local tick
local wait_status
if [[ "$group" =~ ^[1-9][0-9]*$ ]]; then
kill -TERM -- "-$group" 2>/dev/null || :
for tick in {1..5}; do
for _ in {1..5}; do
group_alive "$group" || break
/bin/sleep 0.02
/bin/sleep 0.02 || :
done
if group_alive "$group"; then
kill -KILL -- "-$group" 2>/dev/null || :
fi
fi
if [[ "$leader" =~ ^[1-9][0-9]*$ ]]; then
wait "$leader" 2>/dev/null || :
while :; do
wait_status=0
wait "$leader" 2>/dev/null || wait_status=$?
case "$wait_status" in
129|130|143)
kill -0 "$leader" 2>/dev/null && continue
break
;;
*) break ;;
esac
done
fi
ACTIVE_CHILD_GROUP=''
ACTIVE_CHILD_PID=''
}
signal_exit() {
local status=${1:-1}
trap '' HUP INT TERM
trap - EXIT
exec >/dev/null 2>&1
terminate_active_group
cleanup
if ! cleanup; then
exit 1
fi
exit "$status"
}
handle_signal() {
Expand All @@ -84,6 +104,10 @@ trap 'handle_signal 129' HUP
trap 'handle_signal 130' INT
trap 'handle_signal 143' TERM
/bin/mkdir -m 700 "$run_tmp/home"
child_notify="$run_tmp/child.notify"
/usr/bin/mkfifo "$child_notify" || runner_error
exec 4<> "$child_notify" || runner_error
/bin/rm -f -- "$child_notify" || runner_error

sha_file() { /usr/bin/shasum -a 256 "$1" | /usr/bin/awk '{print $1}'; }
sha_text() { /usr/bin/printf '%s' "$1" | /usr/bin/shasum -a 256 | /usr/bin/awk '{print $1}'; }
Expand Down Expand Up @@ -352,14 +376,16 @@ run_child() {
local diagnostic=$5
local child_home=$6
local dependency=${7:-}
local pid tick=0 pending_status
local pid pending_status child_token notification='' active_marker
CHILD_STATUS=0
CHILD_ERROR=''
/bin/mkdir -m 700 "$child_home"
if [ -n "${ACTIVE_CHILD_GROUP:-}" ] || [ -n "${ACTIVE_CHILD_PID:-}" ]; then
CHILD_ERROR=E_RUNTIME
return 1
fi
CHILD_SEQUENCE=$((CHILD_SEQUENCE + 1))
child_token=$CHILD_SEQUENCE
PENDING_SIGNAL_STATUS=0
SIGNAL_DEFER=1
set -m
Expand All @@ -368,13 +394,19 @@ run_child() {
child_args=("$request" "$jq_bin" "$contract" "$core_fixture_modules")
[ -z "$dependency" ] || child_args+=("$dependency")
[ "$mode" = direct ] || child_args=("$mode" "${child_args[@]}")
exec /usr/bin/env -i HOME="$child_home" TMPDIR="$child_home" PATH=/usr/bin:/bin \
LC_ALL=C GIT_CONFIG_NOSYSTEM=1 GIT_CONFIG_GLOBAL=/dev/null \
GIT_NO_REPLACE_OBJECTS=1 GIT_NO_LAZY_FETCH=1 GIT_TERMINAL_PROMPT=0 \
GIT_AUTHOR_NAME=fake GIT_AUTHOR_EMAIL=fake@example.invalid \
GIT_COMMITTER_NAME=fake GIT_COMMITTER_EMAIL=fake@example.invalid \
GIT_AUTHOR_DATE=2000-01-01T00:00:00Z GIT_COMMITTER_DATE=2000-01-01T00:00:00Z \
/bin/bash "$executable" "${child_args[@]}"
child_status=0
(
exec 4>&-
exec /usr/bin/env -i HOME="$child_home" TMPDIR="$child_home" PATH=/usr/bin:/bin \
LC_ALL=C GIT_CONFIG_NOSYSTEM=1 GIT_CONFIG_GLOBAL=/dev/null \
GIT_NO_REPLACE_OBJECTS=1 GIT_NO_LAZY_FETCH=1 GIT_TERMINAL_PROMPT=0 \
GIT_AUTHOR_NAME=fake GIT_AUTHOR_EMAIL=fake@example.invalid \
GIT_COMMITTER_NAME=fake GIT_COMMITTER_EMAIL=fake@example.invalid \
GIT_AUTHOR_DATE=2000-01-01T00:00:00Z GIT_COMMITTER_DATE=2000-01-01T00:00:00Z \
/bin/bash "$executable" "${child_args[@]}"
) || child_status=$?
printf '%s\n' "$child_token" >&4 || :
exit "$child_status"
) > "$output" 2> "$diagnostic" &
pid=$!
ACTIVE_CHILD_GROUP=$pid
Expand All @@ -387,15 +419,21 @@ run_child() {
SIGNAL_EXITING=1
signal_exit "$pending_status"
fi
while kill -0 "$pid" 2>/dev/null; do
tick=$((tick + 1))
if [ "$tick" -ge 20 ]; then
terminate_active_group
CHILD_ERROR=E_TIMEOUT
return 1
fi
/bin/sleep 0.05
active_marker="$child_home/.runner-active-group"
if ! printf '%s\n' "$pid" > "$active_marker"; then
terminate_active_group
CHILD_ERROR=E_RUNTIME
return 1
fi
while IFS= read -r -t 1 -u 4 notification; do
[ "$notification" = "$child_token" ] && break
notification=''
done
if [ "$notification" != "$child_token" ] && kill -0 "$pid" 2>/dev/null; then
terminate_active_group
CHILD_ERROR=E_TIMEOUT
return 1
fi
wait "$pid" || CHILD_STATUS=$?
if group_alive "${ACTIVE_CHILD_GROUP:-}"; then
terminate_active_group
Expand Down
44 changes: 38 additions & 6 deletions scripts/test/portable-adapter-contracts.test.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -379,12 +379,34 @@ pass 'late target commit recheck'

signal_out="$tmp/signal.out"
signal_err="$tmp/signal.err"
set -m
PATH="$bin:/usr/bin:/bin" "$runner" "$inventory" "$fixture" > "$signal_out" 2> "$signal_err" &
signal_pid=$!
signal_groups=$(jobs -p)
set +m
signal_group_count=$(/usr/bin/printf '%s\n' "$signal_groups" |
/usr/bin/awk 'NF { count++ } END { print count + 0 }')
signal_group=$signal_groups
if [ "$signal_group_count" -ne 1 ] || ! [[ "$signal_group" =~ ^[1-9][0-9]*$ ]] ||
[ "$signal_group" != "$signal_pid" ]; then
/bin/kill "$signal_pid" 2>/dev/null || :
wait "$signal_pid" 2>/dev/null || :
fail 'isolated signal process group'
fi
signal_marker=''
signal_child_group=''
for _ in {1..6000}; do
signal_marker=$(/usr/bin/find "$fixture/scratch" -name 'request-payload.1' -type f -print -quit)
[ -z "$signal_marker" ] || break
signal_marker=$(/usr/bin/find "$fixture/scratch" -path '*/timeout.home/.runner-active-group' \
-type f -print -quit)
if [ -n "$signal_marker" ]; then
signal_child_group=''
IFS= read -r signal_child_group < "$signal_marker" || :
if [[ "$signal_child_group" =~ ^[1-9][0-9]*$ ]] &&
kill -0 -- "-$signal_child_group" 2>/dev/null; then
break
fi
signal_marker=''
fi
/bin/kill -0 "$signal_pid" 2>/dev/null || break
/bin/sleep 0.02
done
Expand All @@ -396,7 +418,7 @@ fi
signal_sent=0
for _ in {1..200}; do
/bin/kill -0 "$signal_pid" 2>/dev/null || break
/bin/kill -TERM "$signal_pid" 2>/dev/null || break
kill -TERM -- "-$signal_group" 2>/dev/null || break
signal_sent=$((signal_sent + 1))
/bin/sleep 0.005
done
Expand All @@ -405,13 +427,22 @@ wait "$signal_pid" || signal_status=$?
/bin/sleep 2.2
signal_stdout_bytes=$(/usr/bin/wc -c < "$signal_out" | /usr/bin/tr -d ' ')
signal_stderr_bytes=$(/usr/bin/wc -c < "$signal_err" | /usr/bin/tr -d ' ')
signal_stderr_sha256=$(sha_file "$signal_err")
signal_stdout_class=empty
signal_stderr_class=empty
[ "$signal_stdout_bytes" -eq 0 ] || signal_stdout_class=other
if [ "$signal_stderr_bytes" -ne 0 ]; then
if /usr/bin/grep -Eq '(^|[[:space:]])(Terminated|Killed|Hangup|Interrupt|Done)(:|[[:space:]]|$)' \
"$signal_err"; then
signal_stderr_class=job-control
elif /usr/bin/grep -Eq 'warning: run_pending_traps: bad value in trap_list\[[0-9]+\]: 0x[[:xdigit:]]+' \
"$signal_err"; then
signal_stderr_class=bash-trap-table
elif /usr/bin/grep -Eq '(^|[[:space:]])E_[A-Z_]+([[:space:]]|$)' "$signal_err"; then
signal_stderr_class=typed-error
elif /usr/bin/grep -Eq '(line [0-9]+:|unbound variable|not a child|syntax error)' \
"$signal_err"; then
signal_stderr_class=shell-runtime
else
signal_stderr_class=other
fi
Expand All @@ -422,12 +453,13 @@ if /usr/bin/find "$fixture/scratch" \( -name timeout-survived -o -name descendan
-print -quit | /usr/bin/grep -q .; then
signal_survivor=present
fi
if [ "$signal_sent" -le 0 ] || [ "$signal_status" -ne 143 ] ||
if [ "$signal_sent" -lt 2 ] || [ "$signal_status" -ne 143 ] ||
[ "$signal_stdout_bytes" -ne 0 ] || [ "$signal_stderr_bytes" -ne 0 ] ||
[ "$signal_scratch_count" -ne 0 ] || [ "$signal_survivor" != absent ]; then
/usr/bin/printf 'signal-debug status=%s sent=%s stdout_bytes=%s stdout_class=%s stderr_bytes=%s stderr_class=%s scratch_entries=%s survivor=%s\n' \
/usr/bin/printf 'signal-debug status=%s sent=%s stdout_bytes=%s stdout_class=%s stderr_bytes=%s stderr_sha256=%s stderr_class=%s scratch_entries=%s survivor=%s\n' \
"$signal_status" "$signal_sent" "$signal_stdout_bytes" "$signal_stdout_class" \
"$signal_stderr_bytes" "$signal_stderr_class" "$signal_scratch_count" "$signal_survivor" >&2
"$signal_stderr_bytes" "$signal_stderr_sha256" "$signal_stderr_class" \
"$signal_scratch_count" "$signal_survivor" >&2
fail 'signal group cleanup'
fi
pass 'TERM stops adapter group and cleans scratch'
Expand Down