fix(mosaic): harden Gitea pr merge fallback #521

Open
jason.woltje wants to merge 2 commits from fix/t_301e4e3b-pr-merge-gitea-empty-uid into main
4 changed files with 424 additions and 6 deletions

View File

@@ -0,0 +1,31 @@
# Scratchpad: t_301e4e3b pr-merge.sh Gitea empty-uid fallback
## Task
Implement a narrow hardening in `packages/mosaic/framework/tools/git/pr-merge.sh` so Gitea merges recover from the known non-interactive `tea pr merge` identity failure: `user does not exist [uid: 0, name: ]`.
## Constraints
- Preserve Mosaic policy gates: squash-only, base branch `main`, queue guard unless explicitly skipped.
- Preserve the existing authenticated Gitea API fallback when no tea login exists.
- Do not fallback on arbitrary tea failures.
- Do not expose tokens or credential-bearing remotes.
- Scope is limited to the merge wrapper plus focused test/support/scratchpad files.
## External issue
- Gitea issue #520: Harden pr-merge.sh Gitea empty-uid fallback
## Plan
1. Add a focused shell regression harness with mocked `tea` and `curl` proving the known empty uid/name failure must fall back to Gitea API.
2. Watch the harness fail on current code.
3. Implement helper functions in `pr-merge.sh` for redacted command display, known failure classification, and authenticated Gitea API merge fallback.
4. Keep unknown `tea` failures blocking by replaying stderr and exiting non-zero.
5. Run syntax, shellcheck if available, focused regression, and repo quality gates before push/PR.
## Session log
- 2026-05-22: Read Kanban context, Mosaic global/repo instructions, created isolated branch `fix/t_301e4e3b-pr-merge-gitea-empty-uid`, and opened Gitea issue #520 using the Mosaic issue wrapper/API fallback.
- 2026-05-22: Added regression harness and watched it fail on current behavior with `user does not exist [uid: 0, name: ]`; implemented narrow fallback and verified known-empty-identity fallback, arbitrary tea failure blocking, and no-tea-login API fallback paths.
- 2026-05-22: Validation passed for `bash -n`, `shellcheck -x`, focused shell harness, `pnpm typecheck`, `pnpm lint`, `pnpm format:check`, and `pnpm --filter @mosaicstack/mosaic test`. Full `pnpm test` exposed an out-of-scope gateway DB setup failure (`relation "messages" does not exist`) in `apps/gateway/src/__tests__/cross-user-isolation.test.ts`.

View File

@@ -0,0 +1,48 @@
# t_5aab9cc8 — pr-merge.sh eval injection remediation
## Objective
Remediate PR #521 review blocker: `packages/mosaic/framework/tools/git/pr-merge.sh` must reject non-numeric PR numbers before metadata lookup/merge and must not use `eval` for GitHub merge execution.
## Scope
- Shell wrapper only: `packages/mosaic/framework/tools/git/pr-merge.sh`
- Focused regression harness: `packages/mosaic/framework/tools/git/test-pr-merge-gitea-empty-uid.sh`
- No API/frontend/infra surfaces.
## Acceptance Criteria
- AC1: `PR_NUMBER` is validated as digits-only immediately after required-argument parsing, before metadata lookup.
- AC2: GitHub merge path uses a quoted argv array, not command-string construction plus `eval`.
- AC3: Focused tests prove PR-number metacharacters are rejected and cannot execute injected shell commands on GitHub path.
- AC4: Focused tests prove PR-number metacharacters are rejected on Gitea path before tea/curl merge calls.
- AC5: Existing Gitea empty-uid fallback behavior remains green.
- AC6: Syntax, shellcheck where available, focused harness, and relevant repo gates are rerun or absence documented.
## Plan
1. Add failing regression tests for GitHub eval injection and Gitea invalid PR rejection.
2. Implement fail-closed PR number validation before metadata lookup.
3. Replace GitHub `eval` command with argv array execution.
4. Run required validation and update this scratchpad with evidence.
5. Commit, queue-guard, push branch, update PR #521.
## TDD Log
- RED: `AGENT_WORK_ROOT="$HERMES_KANBAN_WORKSPACE/work" bash packages/mosaic/framework/tools/git/test-pr-merge-gitea-empty-uid.sh` failed on vulnerable code with `Expected GitHub metacharacter PR number to be rejected` and showed the injected PR number reached the GitHub merge path.
- GREEN: Added digits-only validation before metadata lookup and replaced GitHub `eval` with an argv array. The focused harness now passes and verifies invalid PR numbers are rejected before GitHub `gh` calls and before Gitea `tea`/`curl` calls.
## Validation Evidence
- PASS: `AGENT_WORK_ROOT="$HERMES_KANBAN_WORKSPACE/work" bash -n packages/mosaic/framework/tools/git/pr-merge.sh packages/mosaic/framework/tools/git/test-pr-merge-gitea-empty-uid.sh`
- PASS: `shellcheck -x packages/mosaic/framework/tools/git/pr-merge.sh packages/mosaic/framework/tools/git/test-pr-merge-gitea-empty-uid.sh`
- PASS: `AGENT_WORK_ROOT="$HERMES_KANBAN_WORKSPACE/work" bash packages/mosaic/framework/tools/git/test-pr-merge-gitea-empty-uid.sh`
- PASS: `pnpm --filter @mosaicstack/mosaic... build`
- PASS: `pnpm --filter @mosaicstack/mosaic lint`
- PASS: `pnpm --filter @mosaicstack/mosaic typecheck`
- PASS: `pnpm --filter @mosaicstack/mosaic test` — 32 files / 291 tests passed.
- REVIEW: `/home/hermes/.config/mosaic/tools/codex/codex-code-review.sh --uncommitted` could not run due Codex 401 Unauthorized. Independent delegate review completed read-only with PASS / no blockers; non-blocking suggestion to assert GitHub mock log remains empty was applied.
## Risks / Blockers
- No active blockers.

View File

@@ -2,9 +2,10 @@
# pr-merge.sh - Merge pull requests on Gitea or GitHub
# Usage: pr-merge.sh -n PR_NUMBER [-m squash] [-d] [--skip-queue-guard]
set -e
set -euo pipefail
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
# shellcheck source=packages/mosaic/framework/tools/git/detect-platform.sh
source "$SCRIPT_DIR/detect-platform.sh"
# Default values
@@ -69,6 +70,11 @@ if [[ -z "$PR_NUMBER" ]]; then
usage
fi
if [[ ! "$PR_NUMBER" =~ ^[0-9]+$ ]]; then
echo "Error: Invalid PR number '$PR_NUMBER'. PR number must contain digits only." >&2
exit 1
fi
if [[ "$MERGE_METHOD" != "squash" ]]; then
echo "Error: Mosaic policy enforces squash merge only. Received '$MERGE_METHOD'." >&2
exit 1
@@ -92,21 +98,138 @@ PLATFORM=$(detect_platform)
OWNER=$(get_repo_owner)
REPO=$(get_repo_name)
find_tea_login_for_host() {
local host="$1"
local logins_json
command -v tea >/dev/null 2>&1 || return 1
logins_json=$(tea login list --output json 2>/dev/null) || return 1
TEA_LOGINS_JSON="$logins_json" python3 - "$host" <<'PY'
import json
import os
import sys
host = sys.argv[1]
try:
logins = json.loads(os.environ.get("TEA_LOGINS_JSON", "[]"))
except Exception:
raise SystemExit(1)
for login in logins if isinstance(logins, list) else []:
url = str(login.get("url") or login.get("URL") or "")
name = str(login.get("name") or login.get("Name") or "")
if url.rstrip("/").endswith(host) and name:
print(name)
raise SystemExit(0)
raise SystemExit(1)
PY
}
is_known_tea_empty_identity_failure() {
local error_file="$1"
python3 - "$error_file" <<'PY'
import re
import sys
with open(sys.argv[1], encoding="utf-8", errors="replace") as handle:
error = handle.read()
known_empty_identity = re.search(
r"user does not exist.*\[.*uid:\s*0,\s*name:\s*\]",
error,
flags=re.IGNORECASE | re.DOTALL,
)
raise SystemExit(0 if known_empty_identity else 1)
PY
}
merge_gitea_with_api() {
local host="$1"
local api_url="https://${host}/api/v1/repos/${OWNER}/${REPO}/pulls/${PR_NUMBER}/merge"
local token body_file payload
token=$(get_gitea_token "$host" || true)
if [[ -z "$token" ]]; then
echo "Error: No Gitea API token available for authenticated merge fallback on $host." >&2
return 1
fi
mkdir -p "${AGENT_WORK_ROOT:-/home/hermes/agent-work}"
body_file=$(mktemp "${AGENT_WORK_ROOT:-/home/hermes/agent-work}/pr-merge-api-response.XXXXXX")
payload='{"Do":"squash"}'
if curl -fsS \
-X POST \
-H "Authorization: token $token" \
-H 'Content-Type: application/json' \
-d "$payload" \
"$api_url" > "$body_file"; then
rm -f "$body_file"
return 0
fi
python3 - "$body_file" <<'PY' >&2
import json
import sys
path = sys.argv[1]
try:
with open(path, encoding="utf-8", errors="replace") as handle:
raw = handle.read(500)
data = json.loads(raw) if raw else {}
message = data.get("message") or data.get("error") or raw or "empty response"
except Exception:
try:
with open(path, encoding="utf-8", errors="replace") as handle:
message = handle.read(500) or "empty response"
except Exception:
message = "unreadable response"
print(f"Error: Gitea API merge fallback failed: {message}")
PY
rm -f "$body_file"
return 1
}
case "$PLATFORM" in
github)
CMD="gh pr merge $PR_NUMBER --squash"
[[ "$DELETE_BRANCH" == true ]] && CMD="$CMD --delete-branch"
eval "$CMD"
cmd=(gh pr merge "$PR_NUMBER" --squash)
[[ "$DELETE_BRANCH" == true ]] && cmd+=(--delete-branch)
"${cmd[@]}"
;;
gitea)
CMD="tea pr merge $PR_NUMBER --style squash --repo $OWNER/$REPO --login ${GITEA_LOGIN:-mosaicstack}"
HOST=$(get_remote_host) || {
echo "Error: Cannot determine host from origin remote URL" >&2
exit 1
}
TEA_LOGIN="${GITEA_LOGIN:-$(find_tea_login_for_host "$HOST" || true)}"
# Delete branch after merge if requested
if [[ "$DELETE_BRANCH" == true ]]; then
echo "Note: Branch deletion after merge may need to be done separately with tea" >&2
fi
eval "$CMD"
if [[ -n "$TEA_LOGIN" ]]; then
mkdir -p "${AGENT_WORK_ROOT:-/home/hermes/agent-work}"
TEA_ERROR_FILE=$(mktemp "${AGENT_WORK_ROOT:-/home/hermes/agent-work}/pr-merge-tea-error.XXXXXX")
if tea pr merge "$PR_NUMBER" --style squash --repo "$OWNER/$REPO" --login "$TEA_LOGIN" 2> "$TEA_ERROR_FILE"; then
rm -f "$TEA_ERROR_FILE"
elif is_known_tea_empty_identity_failure "$TEA_ERROR_FILE"; then
cat "$TEA_ERROR_FILE" >&2
echo "Known tea empty identity failure detected; using authenticated Gitea API merge fallback." >&2
rm -f "$TEA_ERROR_FILE"
merge_gitea_with_api "$HOST"
else
cat "$TEA_ERROR_FILE" >&2
rm -f "$TEA_ERROR_FILE"
exit 1
fi
else
echo "No tea login configured for $HOST; using authenticated Gitea API merge fallback." >&2
merge_gitea_with_api "$HOST"
fi
;;
*)
echo "Error: Could not detect git platform" >&2

View File

@@ -0,0 +1,216 @@
#!/bin/bash
# Regression harness for pr-merge.sh Gitea non-interactive tea empty identity fallback.
set -euo pipefail
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
WORK_ROOT="${AGENT_WORK_ROOT:-/home/hermes/agent-work}"
SANDBOX="$WORK_ROOT/pr-merge-empty-uid-test-$$"
MOCK_BIN="$SANDBOX/bin"
REPO_DIR="$SANDBOX/repo"
LOG_FILE="$SANDBOX/mock.log"
cleanup() {
rm -rf "$SANDBOX"
}
trap cleanup EXIT
mkdir -p "$MOCK_BIN" "$REPO_DIR"
: > "$LOG_FILE"
cat > "$MOCK_BIN/tea" <<'EOF'
#!/bin/bash
set -euo pipefail
printf 'tea %q ' "$@" >> "$PR_MERGE_TEST_LOG"
printf '\n' >> "$PR_MERGE_TEST_LOG"
if [[ "$*" == *"pr merge"* ]]; then
echo 'user does not exist [uid: 0, name: ]' >&2
exit 1
fi
exit 0
EOF
chmod +x "$MOCK_BIN/tea"
cat > "$MOCK_BIN/curl" <<'EOF'
#!/bin/bash
set -euo pipefail
printf 'curl %q ' "$@" >> "$PR_MERGE_TEST_LOG"
printf '\n' >> "$PR_MERGE_TEST_LOG"
args=" $* "
if [[ "$args" == *"/api/v1/repos/mosaicstack/stack/pulls/123"* && "$args" != *"/api/v1/repos/mosaicstack/stack/pulls/123/merge"* ]]; then
cat <<'JSON'
{"number":123,"title":"mock","state":"open","user":{"login":"tester"},"head":{"ref":"feature/mock"},"base":{"ref":"main"},"labels":[],"assignees":[],"html_url":"https://git.mosaicstack.dev/mosaicstack/stack/pulls/123","mergeable":true}
JSON
exit 0
fi
if [[ "$args" == *"-X POST"* && "$args" == *"/api/v1/repos/mosaicstack/stack/pulls/123/merge"* ]]; then
cat <<'JSON'
{"merged":true,"message":"mock merge complete"}
JSON
exit 0
fi
echo "unexpected curl invocation: $*" >&2
exit 97
EOF
chmod +x "$MOCK_BIN/curl"
cd "$REPO_DIR"
git init -q
git remote add origin https://git.mosaicstack.dev/mosaicstack/stack.git
export PATH="$MOCK_BIN:$PATH"
export PR_MERGE_TEST_LOG="$LOG_FILE"
export GITEA_LOGIN="git.mosaicstack.dev"
export GITEA_TOKEN="redacted-test-token"
OUTPUT="$SANDBOX/output.log"
if ! "$SCRIPT_DIR/pr-merge.sh" -n 123 -m squash --skip-queue-guard > "$OUTPUT" 2>&1; then
echo "Expected pr-merge.sh to recover via Gitea API fallback." >&2
echo "--- output ---" >&2
sed 's/redacted-test-token/***REDACTED***/g' "$OUTPUT" >&2
echo "--- mock log ---" >&2
sed 's/redacted-test-token/***REDACTED***/g' "$LOG_FILE" >&2
exit 1
fi
if ! grep -q '/api/v1/repos/mosaicstack/stack/pulls/123/merge' "$LOG_FILE"; then
echo "Expected authenticated Gitea merge API endpoint to be called." >&2
sed 's/redacted-test-token/***REDACTED***/g' "$LOG_FILE" >&2
exit 1
fi
if grep -q 'redacted-test-token' "$OUTPUT"; then
echo "Token leaked to pr-merge.sh output." >&2
exit 1
fi
cat > "$MOCK_BIN/tea" <<'EOF'
#!/bin/bash
set -euo pipefail
printf 'tea %q ' "$@" >> "$PR_MERGE_TEST_LOG"
printf '\n' >> "$PR_MERGE_TEST_LOG"
if [[ "$*" == *"pr merge"* ]]; then
echo 'tea network timeout' >&2
exit 2
fi
exit 0
EOF
chmod +x "$MOCK_BIN/tea"
: > "$LOG_FILE"
if "$SCRIPT_DIR/pr-merge.sh" -n 123 -m squash --skip-queue-guard > "$OUTPUT" 2>&1; then
echo "Expected arbitrary tea failure to remain blocking." >&2
exit 1
fi
if grep -q '/api/v1/repos/mosaicstack/stack/pulls/123/merge' "$LOG_FILE"; then
echo "Arbitrary tea failure unexpectedly used Gitea API merge fallback." >&2
sed 's/redacted-test-token/***REDACTED***/g' "$LOG_FILE" >&2
exit 1
fi
if ! grep -q 'tea network timeout' "$OUTPUT"; then
echo "Expected arbitrary tea error to be preserved in output." >&2
sed 's/redacted-test-token/***REDACTED***/g' "$OUTPUT" >&2
exit 1
fi
cat > "$MOCK_BIN/tea" <<'EOF'
#!/bin/bash
set -euo pipefail
printf 'tea %q ' "$@" >> "$PR_MERGE_TEST_LOG"
printf '\n' >> "$PR_MERGE_TEST_LOG"
if [[ "$*" == *"login list"* ]]; then
echo '[]'
exit 0
fi
if [[ "$*" == *"pr merge"* ]]; then
echo 'tea merge should not run without a configured host login' >&2
exit 99
fi
exit 0
EOF
chmod +x "$MOCK_BIN/tea"
unset GITEA_LOGIN
: > "$LOG_FILE"
if ! "$SCRIPT_DIR/pr-merge.sh" -n 123 -m squash --skip-queue-guard > "$OUTPUT" 2>&1; then
echo "Expected missing tea login to use authenticated Gitea API fallback." >&2
sed 's/redacted-test-token/***REDACTED***/g' "$OUTPUT" >&2
sed 's/redacted-test-token/***REDACTED***/g' "$LOG_FILE" >&2
exit 1
fi
if ! grep -q '/api/v1/repos/mosaicstack/stack/pulls/123/merge' "$LOG_FILE"; then
echo "Expected missing tea login path to call Gitea API merge endpoint." >&2
sed 's/redacted-test-token/***REDACTED***/g' "$LOG_FILE" >&2
exit 1
fi
SENTINEL="$SANDBOX/injected-sentinel"
INJECTION="123; touch $SENTINEL #"
cat > "$MOCK_BIN/gh" <<'EOF'
#!/bin/bash
set -euo pipefail
printf 'gh %q ' "$@" >> "$PR_MERGE_TEST_LOG"
printf '\n' >> "$PR_MERGE_TEST_LOG"
if [[ "$*" == *"pr view"* ]]; then
cat <<'JSON'
{"number":123,"title":"mock","baseRefName":"main","headRefName":"feature/mock"}
JSON
exit 0
fi
if [[ "$*" == *"pr merge"* ]]; then
exit 0
fi
echo "unexpected gh invocation: $*" >&2
exit 98
EOF
chmod +x "$MOCK_BIN/gh"
cd "$REPO_DIR"
git remote set-url origin https://github.com/mosaicstack/stack.git
: > "$LOG_FILE"
rm -f "$SENTINEL"
if "$SCRIPT_DIR/pr-merge.sh" -n "$INJECTION" -m squash --skip-queue-guard > "$OUTPUT" 2>&1; then
echo "Expected GitHub metacharacter PR number to be rejected." >&2
sed 's/redacted-test-token/***REDACTED***/g' "$OUTPUT" >&2
exit 1
fi
if [[ -e "$SENTINEL" ]]; then
echo "GitHub metacharacter PR number executed injected shell command." >&2
exit 1
fi
if [[ -s "$LOG_FILE" ]]; then
echo "GitHub metacharacter PR number should be rejected before gh calls." >&2
sed 's/redacted-test-token/***REDACTED***/g' "$LOG_FILE" >&2
exit 1
fi
if ! grep -q 'Invalid PR number' "$OUTPUT"; then
echo "Expected invalid PR number error for GitHub metacharacter input." >&2
sed 's/redacted-test-token/***REDACTED***/g' "$OUTPUT" >&2
exit 1
fi
cd "$REPO_DIR"
git remote set-url origin https://git.mosaicstack.dev/mosaicstack/stack.git
export GITEA_LOGIN="git.mosaicstack.dev"
: > "$LOG_FILE"
rm -f "$SENTINEL"
if "$SCRIPT_DIR/pr-merge.sh" -n "$INJECTION" -m squash --skip-queue-guard > "$OUTPUT" 2>&1; then
echo "Expected Gitea metacharacter PR number to be rejected." >&2
sed 's/redacted-test-token/***REDACTED***/g' "$OUTPUT" >&2
exit 1
fi
if [[ -e "$SENTINEL" ]]; then
echo "Gitea metacharacter PR number executed injected shell command." >&2
exit 1
fi
if [[ -s "$LOG_FILE" ]]; then
echo "Gitea metacharacter PR number should be rejected before tea/curl calls." >&2
sed 's/redacted-test-token/***REDACTED***/g' "$LOG_FILE" >&2
exit 1
fi
if ! grep -q 'Invalid PR number' "$OUTPUT"; then
echo "Expected invalid PR number error for Gitea metacharacter input." >&2
sed 's/redacted-test-token/***REDACTED***/g' "$OUTPUT" >&2
exit 1
fi
echo "pr-merge.sh Gitea fallback regression passed"