mirror of
https://github.com/anthropics/claude-plugins-official.git
synced 2026-09-09 02:02:03 -03:00
a310cc1ae6
Three stop-hook bugs that break loop termination:
- A bare assistant message equal to the promise (e.g. just "DONE") ended
the loop even without <promise> tags: the perl substitution passed the
input through unchanged when the tag was absent. Now the promise text
is only extracted when a <promise>...</promise> tag is present.
- The observed promise text was whitespace-normalized but the expected
promise was not, so a promise containing consecutive spaces
("ALL DONE") could never match and the loop never ended. Both sides
are now normalized identically (trim + collapse runs of whitespace).
- The iteration reader accepted "iteration:1" (no space after the colon)
but the writer's sed only matched "iteration: " with a space, so the
counter froze and the loop ran past max_iterations. The writer now
matches with or without spaces, like the reader.
Adds tests/stop-hook.test.sh covering all three regressions plus
controls; the three bug cases fail on the previous script and pass now.
Fixes anthropics/claude-code#81827
Fixes anthropics/claude-code#81828
Fixes anthropics/claude-code#81829
No-Verification-Needed: shell-script plugin in claude-plugins-official; no claude-cli-internal runtime surface
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
113 lines
4.3 KiB
Bash
Executable File
113 lines
4.3 KiB
Bash
Executable File
#!/bin/bash
|
|
|
|
# Regression tests for hooks/stop-hook.sh
|
|
# Run: bash plugins/ralph-loop/tests/stop-hook.test.sh
|
|
#
|
|
# Covers:
|
|
# 1. A bare completion word WITHOUT <promise> tags must NOT end the loop
|
|
# (anthropics/claude-code#81827)
|
|
# 2. A promise containing multiple spaces must still match a tagged promise
|
|
# (whitespace is normalized on BOTH sides) (anthropics/claude-code#81828)
|
|
# 3. "iteration:N" (no space after the colon) must still be advanced by the
|
|
# writer, so max_iterations terminates the loop (anthropics/claude-code#81829)
|
|
# 4. Control: a normal in-progress message blocks the stop and bumps iteration
|
|
|
|
set -uo pipefail
|
|
|
|
HOOK="$(cd "$(dirname "$0")/.." && pwd)/hooks/stop-hook.sh"
|
|
FAILURES=0
|
|
|
|
# run_hook <workdir> <last-assistant-text>
|
|
# Prints the hook's stdout; hook exit status is propagated.
|
|
run_hook() {
|
|
local workdir=$1 text=$2
|
|
local transcript="$workdir/transcript.jsonl"
|
|
jq -cn --arg t "$text" '{"role":"assistant","message":{"content":[{"type":"text","text":$t}]}}' > "$transcript"
|
|
(cd "$workdir" && jq -cn --arg p "$transcript" '{"session_id":"test-session","transcript_path":$p}' | bash "$HOOK")
|
|
}
|
|
|
|
# write_state <workdir> <iteration-line> <max_iterations> <completion_promise-yaml>
|
|
write_state() {
|
|
local workdir=$1 iteration_line=$2 max=$3 promise=$4
|
|
mkdir -p "$workdir/.claude"
|
|
cat > "$workdir/.claude/ralph-loop.local.md" <<EOF
|
|
---
|
|
active: true
|
|
$iteration_line
|
|
session_id: test-session
|
|
max_iterations: $max
|
|
completion_promise: $promise
|
|
started_at: "2026-01-01T00:00:00Z"
|
|
---
|
|
|
|
keep working on the task
|
|
EOF
|
|
}
|
|
|
|
check() {
|
|
local name=$1 ok=$2
|
|
if [[ "$ok" == "true" ]]; then
|
|
echo "PASS: $name"
|
|
else
|
|
echo "FAIL: $name"
|
|
FAILURES=$((FAILURES + 1))
|
|
fi
|
|
}
|
|
|
|
TMP=$(mktemp -d "${TMPDIR:-/tmp}/ralph-stop-hook-test.XXXXXX") || { echo "mktemp failed" >&2; exit 1; }
|
|
trap 'rm -rf "$TMP"' EXIT
|
|
|
|
# Case 1 (#81827): bare "DONE" with no <promise> tag must NOT complete the loop.
|
|
d="$TMP/case1"; mkdir -p "$d"
|
|
write_state "$d" "iteration: 1" 0 '"DONE"'
|
|
out=$(run_hook "$d" "DONE")
|
|
check "untagged bare word does not complete" \
|
|
"$([[ -f "$d/.claude/ralph-loop.local.md" ]] && grep -q '"decision": "block"' <<< "$out" && echo true || echo false)"
|
|
|
|
# Case 1b: properly tagged <promise>DONE</promise> completes and removes state.
|
|
d="$TMP/case1b"; mkdir -p "$d"
|
|
write_state "$d" "iteration: 1" 0 '"DONE"'
|
|
out=$(run_hook "$d" "All finished. <promise>DONE</promise>")
|
|
check "tagged promise completes" \
|
|
"$([[ ! -f "$d/.claude/ralph-loop.local.md" ]] && grep -q "Detected" <<< "$out" && echo true || echo false)"
|
|
|
|
# Case 2 (#81828): promise with a double space must match its tagged emission.
|
|
d="$TMP/case2"; mkdir -p "$d"
|
|
write_state "$d" "iteration: 1" 0 '"ALL DONE"'
|
|
out=$(run_hook "$d" "<promise>ALL DONE</promise>")
|
|
check "double-space promise completes" \
|
|
"$([[ ! -f "$d/.claude/ralph-loop.local.md" ]] && grep -q "Detected" <<< "$out" && echo true || echo false)"
|
|
|
|
# Case 2b: single-space promise still works.
|
|
d="$TMP/case2b"; mkdir -p "$d"
|
|
write_state "$d" "iteration: 1" 0 '"ALL DONE"'
|
|
out=$(run_hook "$d" "<promise>ALL DONE</promise>")
|
|
check "single-space promise completes" \
|
|
"$([[ ! -f "$d/.claude/ralph-loop.local.md" ]] && grep -q "Detected" <<< "$out" && echo true || echo false)"
|
|
|
|
# Case 3 (#81829): "iteration:1" (no space) must still advance, and the loop
|
|
# must terminate once max_iterations is reached.
|
|
d="$TMP/case3"; mkdir -p "$d"
|
|
write_state "$d" "iteration:1" 2 null
|
|
out=$(run_hook "$d" "still working on it")
|
|
advanced=$(grep -c '^iteration: 2$' "$d/.claude/ralph-loop.local.md" || true)
|
|
check "no-space iteration counter advances" \
|
|
"$([[ "$advanced" == "1" ]] && grep -q '"decision": "block"' <<< "$out" && echo true || echo false)"
|
|
out=$(run_hook "$d" "still working on it")
|
|
check "loop terminates at max_iterations" \
|
|
"$([[ ! -f "$d/.claude/ralph-loop.local.md" ]] && grep -q "Max iterations" <<< "$out" && echo true || echo false)"
|
|
|
|
# Case 4 (control): in-progress message blocks and bumps iteration.
|
|
d="$TMP/case4"; mkdir -p "$d"
|
|
write_state "$d" "iteration: 1" 0 '"DONE"'
|
|
out=$(run_hook "$d" "still working on it")
|
|
check "in-progress message blocks and bumps iteration" \
|
|
"$(grep -q '^iteration: 2$' "$d/.claude/ralph-loop.local.md" && grep -q '"decision": "block"' <<< "$out" && echo true || echo false)"
|
|
|
|
echo ""
|
|
if [[ $FAILURES -gt 0 ]]; then
|
|
echo "$FAILURES test(s) failed"
|
|
exit 1
|
|
fi
|
|
echo "All tests passed"
|