redirector.sh: never leave it ambiguous which binary is under test

Two defects, both found by the guards misfiring rather than by reading:

1. BIN preferred target/debug and fell back to release only when debug was
   absent. `cargo build --release` therefore produced a correct binary while
   the script kept inspecting a stale debug one, and the build guard refused
   with a message naming a commit nobody was trying to run. The guard was
   right that something was stale — it just pointed at the wrong artifact.
   Disagreement between the two is now an explicit refusal naming both, with
   OPENFUT_REDIRECTOR_BIN as the deliberate override.

   The refusal is recorded at load and raised only by `verify` and `start`.
   `stop` and `status` must work in any build-tree state: rollback can never
   be blocked by a question about which artifact would have been started.

2. `verify-running` read the stamp file without checking the process still
   existed. The stamp outlives the process, so after a stop it reported on a
   corpse — either "identity OK" or a REFUSAL naming a commit, both implying
   something was running when nothing was.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This commit is contained in:
funman300
2026-08-11 16:20:16 +00:00
parent 096d1c882f
commit aa2679162d
+44 -2
View File
@@ -22,8 +22,36 @@ PORTFILE="$RUNDIR/redirector.port"
# when the branch ref moves) can silently advance past the live process.
STAMPFILE="$RUNDIR/redirector.commit"
LOGFILE="${OPENFUT_REDIRECTOR_LOG:-$RUNDIR/redirector.log}"
BIN="$ROOT/target/debug/openfut-redirector-host"
[[ -x "$BIN" ]] || BIN="$ROOT/target/release/openfut-redirector-host"
# Which artifact is under test must never be ambiguous. Preferring debug and
# falling back to release meant `cargo build --release` could produce a fresh
# binary while this kept launching a stale debug one — the guard then compared
# the WRONG artifact against HEAD and refused with a message naming a commit
# nobody was trying to run. Both present and disagreeing is an error, not a
# preference. Set OPENFUT_REDIRECTOR_BIN to choose deliberately.
DEBUG_BIN="$ROOT/target/debug/openfut-redirector-host"
RELEASE_BIN="$ROOT/target/release/openfut-redirector-host"
BIN_ERR=""
if [[ -n "${OPENFUT_REDIRECTOR_BIN:-}" ]]; then
BIN="$OPENFUT_REDIRECTOR_BIN"
[[ -x "$BIN" ]] || { echo "redirector: OPENFUT_REDIRECTOR_BIN is not executable: $BIN" >&2; exit 1; }
elif [[ -x "$DEBUG_BIN" && -x "$RELEASE_BIN" ]]; then
d="$("$DEBUG_BIN" --identity 2>&1 | sed -nE 's/.*commit=([0-9a-f]+).*/\1/p' | head -1)"
r="$("$RELEASE_BIN" --identity 2>&1 | sed -nE 's/.*commit=([0-9a-f]+).*/\1/p' | head -1)"
if [[ "$d" != "$r" ]]; then
# Recorded, NOT fatal here: `stop` and `status` must keep working no matter
# what state the build tree is in. Rolling back can never be blocked by a
# question about which artifact would have been started.
BIN_ERR="REFUSING — two binaries exist and disagree.
debug $DEBUG_BIN commit=$d
release $RELEASE_BIN commit=$r
Rebuild both, delete one, or set OPENFUT_REDIRECTOR_BIN."
fi
BIN="$RELEASE_BIN"
elif [[ -x "$DEBUG_BIN" ]]; then
BIN="$DEBUG_BIN"
else
BIN="$RELEASE_BIN"
fi
die() { echo "redirector: $*" >&2; exit 1; }
pid_alive() { kill -0 "$1" 2>/dev/null; }
@@ -48,16 +76,21 @@ list_procs() {
return 0
}
# Only the commands that actually run a binary care which one it is.
need_bin() { [[ -z "$BIN_ERR" ]] || die "$BIN_ERR"; }
# The binary prints `commit=<sha>` in its banner; ask it rather than guessing.
stamped_commit() { "$BIN" --identity 2>&1 | sed -nE 's/.*commit=([0-9a-f]+).*/\1/p' | head -1; }
cmd_verify() {
need_bin
local c; c="$(stamped_commit)"
[[ -n "$c" ]] || die "could not read the binary's commit stamp"
"$ROOT/scripts/verify-build-identity.sh" "$c"
}
cmd_start() {
need_bin
[[ -x "$BIN" ]] || die "not built: cargo build -p openfut-redirector-host"
: "${OPENFUT_REDIRECTOR_HOST_PORT:?set OPENFUT_REDIRECTOR_HOST_PORT (no default: runs beside Python)}"
local strays; strays="$(list_procs)"
@@ -115,6 +148,15 @@ cmd_status() {
}
cmd_verify_running() {
# The stamp outlives the process it describes. Without this check the command
# happily reports on a corpse — either "OK" or a REFUSAL naming a commit,
# both implying something is running when nothing is.
local pid=""
[[ -f "$PIDFILE" ]] && pid="$(cat "$PIDFILE" 2>/dev/null)"
if [[ -z "$pid" ]] || ! pid_alive "$pid"; then
echo "REFUSING: nothing is running — the stamp describes a process that has exited." >&2
return 1
fi
[[ -f "$STAMPFILE" ]] || die "no running-process stamp — was it started by this script?"
local running head
running="$(cat "$STAMPFILE")"; head="$(git -C "$ROOT" rev-parse --short=7 HEAD 2>/dev/null)"