aboutsummaryrefslogtreecommitdiff
path: root/scripts
diff options
context:
space:
mode:
authorStefan Boberg <[email protected]>2026-03-24 13:18:12 +0100
committerGitHub Enterprise <[email protected]>2026-03-24 13:18:12 +0100
commitfec7fea085968bb93291638fd6d8c037682595ec (patch)
tree1871b36ead5b8fea3db12a347e658344afedf572 /scripts
parentrefactor hub notifications (#888) (diff)
downloadzen-fec7fea085968bb93291638fd6d8c037682595ec.tar.xz
zen-fec7fea085968bb93291638fd6d8c037682595ec.zip
Linux Crashpad fix (#890)
- **Replace crashpad static-libc++ patch file with `io.replace()` in `on_install`** — The old `.patch` file was fragile (trailing-whitespace stripping on Windows would silently break it). Using `io.replace()` in the xmake build script is more robust and easier to maintain. - **Skip sentry-native `on_test` link check on Linux** — The link test requires `-lc++abi` when building with the UE clang toolchain but adding it unconditionally breaks GCC/libstdc++ builds. The zenserver build itself validates that the library is usable. - **Add `crashpad-test.sh`** — A test script that launches a release zenserver, waits for the health endpoint, then verifies that `crashpad_handler` is running, no `sentry_init` failure was logged, and the handler has no dynamic `libc++.so.1` dependency. - **Add Crashpad Check step to Linux release CI** — Runs `crashpad-test.sh` in the `validate` workflow for release builds to catch crashpad regressions before merge.
Diffstat (limited to 'scripts')
-rwxr-xr-xscripts/test_linux/crashpad-test.sh184
1 files changed, 184 insertions, 0 deletions
diff --git a/scripts/test_linux/crashpad-test.sh b/scripts/test_linux/crashpad-test.sh
new file mode 100755
index 000000000..bd08cefc5
--- /dev/null
+++ b/scripts/test_linux/crashpad-test.sh
@@ -0,0 +1,184 @@
+#!/usr/bin/env bash
+# Verify that crashpad_handler is active when zenserver (release build) starts.
+#
+# This test:
+# 1. Launches zenserver from the release build.
+# 2. Waits for the HTTP health endpoint to become ready.
+# 3. Checks that a crashpad_handler child process is running.
+# 4. Checks that the startup log contains "sentry initialized" (not a failure).
+#
+# Usage:
+# ./scripts/test_linux/crashpad-test.sh [path-to-zenserver]
+#
+# If no path is given, defaults to build/linux/x86_64/release/zenserver
+# relative to the repository root.
+#
+# The test exits 0 on success, 1 on failure.
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+REPO_ROOT="$(cd "$SCRIPT_DIR/../.." && pwd)"
+
+ZENSERVER_BINARY="${1:-$REPO_ROOT/build/linux/x86_64/release/zenserver}"
+CRASHPAD_HANDLER="$(dirname "$ZENSERVER_BINARY")/crashpad_handler"
+PORT=18558 # Use a non-default port so we don't collide with a running zenserver
+DATA_DIR="$(mktemp -d)"
+STDOUT_FILE="$DATA_DIR/stdout.log"
+ZENSERVER_PID=""
+
+RED='\033[0;31m'
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+NC='\033[0m'
+
+PASSED=0
+FAILED=0
+
+pass() {
+ echo -e " ${GREEN}PASS${NC} $1"
+ (( PASSED++ )) || true
+}
+
+fail() {
+ echo -e " ${RED}FAIL${NC} $1"
+ (( FAILED++ )) || true
+}
+
+cleanup() {
+ set +e
+ if [ -n "$ZENSERVER_PID" ] && kill -0 "$ZENSERVER_PID" 2>/dev/null; then
+ kill "$ZENSERVER_PID" 2>/dev/null
+ wait "$ZENSERVER_PID" 2>/dev/null
+ fi
+ rm -rf "$DATA_DIR"
+}
+trap cleanup EXIT
+
+# ── Preflight ────────────────────────────────────────────────────────────────
+
+echo ""
+echo "=============================="
+echo " Crashpad active check"
+echo "=============================="
+echo ""
+
+if [ ! -f "$ZENSERVER_BINARY" ]; then
+ echo -e "${RED}ERROR: zenserver binary not found: $ZENSERVER_BINARY${NC}"
+ echo " Build with: xmake config -m release && xmake build zenserver"
+ exit 1
+fi
+
+if [ ! -f "$CRASHPAD_HANDLER" ]; then
+ echo -e "${RED}ERROR: crashpad_handler not found alongside zenserver: $CRASHPAD_HANDLER${NC}"
+ echo " It should be copied there automatically by the build."
+ exit 1
+fi
+
+echo "zenserver: $ZENSERVER_BINARY"
+echo "crashpad_handler: $CRASHPAD_HANDLER"
+echo "port: $PORT"
+echo "data dir: $DATA_DIR"
+echo ""
+
+# ── Start zenserver ──────────────────────────────────────────────────────────
+
+# zenserver runs in the foreground until SIGINT/SIGTERM. Launch in background
+# so we can poll its health endpoint and inspect child processes.
+"$ZENSERVER_BINARY" \
+ --port="$PORT" \
+ --data-dir="$DATA_DIR" \
+ > "$STDOUT_FILE" 2>&1 &
+ZENSERVER_PID=$!
+
+echo "Started zenserver (pid $ZENSERVER_PID), waiting for health endpoint..."
+
+# ── Wait for health endpoint ─────────────────────────────────────────────────
+
+READY=false
+for i in $(seq 1 40); do
+ if curl -sf "http://localhost:$PORT/health" > /dev/null 2>&1; then
+ READY=true
+ break
+ fi
+ sleep 0.5
+done
+
+if [ "$READY" = false ]; then
+ echo -e "${RED}ERROR: zenserver did not become ready within 20 seconds${NC}"
+ if [ -f "$STDOUT_FILE" ]; then
+ echo ""
+ echo "--- stdout ---"
+ cat "$STDOUT_FILE"
+ fi
+ exit 1
+fi
+
+echo "Server is ready."
+
+# Give the server a moment to finish startup logging (sentry init log
+# message is emitted after the health endpoint comes up).
+sleep 1
+echo ""
+
+# ── Test 1: crashpad_handler process is running for our data dir ─────────────
+#
+# sentry-native starts crashpad_handler with --database pointing at
+# $DATA_DIR/.sentry-native. The process re-parents itself out of zenserver's
+# process tree, so we look for it by its command-line arguments rather than
+# by parent PID.
+
+HANDLER_PID="$(pgrep -f "crashpad_handler.*${DATA_DIR}" 2>/dev/null | head -1 || true)"
+if [ -n "$HANDLER_PID" ]; then
+ pass "crashpad_handler is running (pid $HANDLER_PID) with database in our data dir"
+else
+ fail "crashpad_handler process not found — sentry_init may have failed or sentry is disabled"
+fi
+
+# ── Test 2: No sentry_init failure in startup log ────────────────────────────
+#
+# The "sentry initialized" success message is logged at INFO level under the
+# sentry-sdk log category, which is filtered to Warn by default — so it won't
+# appear in normal stdout. We check for the absence of the failure message
+# instead (which IS at Warn level and would appear if sentry_init failed).
+
+if grep -q "sentry_init returned failure" "$STDOUT_FILE" 2>/dev/null; then
+ ERRMSG="$(grep "sentry_init returned failure" "$STDOUT_FILE" | head -1)"
+ fail "sentry_init reported failure: $ERRMSG"
+else
+ pass "No sentry_init failure message in startup log"
+fi
+
+# ── Test 3: ldd sanity — crashpad_handler must not need libc++.so.1 ──────────
+
+MISSING_LIBCXX="$(ldd "$CRASHPAD_HANDLER" 2>/dev/null | grep "libc++\.so\.1" | grep "not found" || true)"
+if [ -n "$MISSING_LIBCXX" ]; then
+ fail "crashpad_handler has an unsatisfied libc++.so.1 dependency (static linking patch not applied)"
+elif ldd "$CRASHPAD_HANDLER" 2>/dev/null | grep -q "libc++\.so\.1"; then
+ fail "crashpad_handler links libc++.so.1 dynamically — it should be statically linked"
+else
+ pass "crashpad_handler has no dynamic libc++.so.1 dependency"
+fi
+
+# ── Summary ──────────────────────────────────────────────────────────────────
+
+echo ""
+echo "=============================="
+printf " Passed: "
+echo -e "${GREEN}${PASSED}${NC}"
+printf " Failed: "
+if [ "$FAILED" -gt 0 ]; then
+ echo -e "${RED}${FAILED}${NC}"
+else
+ echo -e "${GREEN}${FAILED}${NC}"
+fi
+echo "=============================="
+echo ""
+
+if [ "$FAILED" -gt 0 ]; then
+ if [ -f "$STDOUT_FILE" ]; then
+ echo "--- stdout ---"
+ cat "$STDOUT_FILE"
+ fi
+ exit 1
+fi