test: add immutable Linux late-slope diagnostic tooling

2026-11-10 17:59:15 -05:00
parent 7023285a3e
commit 613369fc54
7 changed files with 1288 additions and 0 deletions
@@ -0,0 +1,70 @@
+#!/usr/bin/env bash
+# Execute the immutable self-contained apphost through the cached Linux VSTest controller.
+set -euo pipefail
+umask 077
+app=ZB.MOM.WW.OtOpcUa.OpcUaServer.IntegrationTests
+test_class="$app.Sdk2LifecycleSoakTests"
+mkdir -p /task/tmp /task/data /task/cli /task/results
+export TMPDIR=/task/tmp XDG_DATA_HOME=/task/data DOTNET_CLI_HOME=/task/cli
+export OTOPCUA_SDK2_SOAK=1 OTOPCUA_SDK2_SOAK_MODE=late-slope OTOPCUA_SDK2_SOAK_SECONDS=7200
+export OTOPCUA_SDK2_SOAK_OUTPUT=/task/resources.jsonl OPCUA_CERTIFICATE_LEAK_TRACKING=1
+export DOTNET_CLI_TELEMETRY_OPTOUT=1 DOTNET_NOLOGO=1
+[[ $(uname -s) == Linux && $(uname -m) == aarch64 && -d /proc/self/fd ]]
+[[ ${OTOPCUA_SOAK_BUILD_SHA:-} =~ ^[0-9a-f]{40}$ ]]
+dotnet --info > /task/controller-runtime.txt
+
+dotnet vstest "/app/$app.dll" \
+    --TestCaseFilter:"FullyQualifiedName~$test_class.Descriptor_probe_" \
+    --ResultsDirectory:/task/results '--Logger:trx;LogFileName=descriptor-controls.trx' > /task/descriptor-controls.log 2>&1
+
+# Verify actual execution before starting the long test. The host parses the complete TRX independently.
+controls=/task/results/descriptor-controls.trx
+counters=$(grep -E '^[[:space:]]*<Counters ' "$controls")
+[[ $(grep -Ec '^[[:space:]]*<Counters ' "$controls") == 1 ]]
+for attribute in total=2 executed=2 passed=2 failed=0 notExecuted=0 error=0 timeout=0 aborted=0; do
+    [[ $counters == *" ${attribute%=*}=\"${attribute#*=}\""* ]]
+done
+[[ $(grep -Ec '^[[:space:]]*<UnitTestResult ' "$controls") == 2 ]]
+for name in Descriptor_probe_observes_an_open_PKI_file_through_a_symlink_alias_and_its_close \
+            Descriptor_probe_times_out_and_reaps_a_process_that_keeps_both_streams_open; do
+    result=$(grep -E '^[[:space:]]*<UnitTestResult ' "$controls" | grep -F "testName=\"$test_class.$name\"")
+    [[ $result == *' outcome="Passed"'* && $result != *$'\n'* ]]
+done
+
+vstest_pid=
+cleanup() {
+    if [[ -n "$vstest_pid" ]]; then kill "$vstest_pid" 2>/dev/null || true; fi
+}
+trap cleanup EXIT
+dotnet vstest "/app/$app.dll" \
+    --TestCaseFilter:"FullyQualifiedName=$test_class.Repeated_lifecycles_return_resources_to_a_steady_baseline" \
+    --ResultsDirectory:/task/results '--Logger:trx;LogFileName=late-slope.trx' > /task/test.log 2>&1 &
+vstest_pid=$!
+
+# The canonical startup deadline is 120 seconds. Allow controller startup before observing its first complete row.
+proof_deadline=$((SECONDS + 150))
+observed=0
+while ((SECONDS < proof_deadline)) && kill -0 "$vstest_pid" 2>/dev/null; do
+    if [[ -s /task/resources.jsonl ]] && IFS= read -r header < /task/resources.jsonl; then
+        [[ $header == *'"kind":"start"'* && $header == *'"mode":"late-slope"'* ]]
+        pid=$(sed -n 's/.*"processId":\([1-9][0-9]*\),.*/\1/p' <<< "$header")
+        [[ $pid =~ ^[1-9][0-9]*$ ]]
+        executable=$(readlink "/proc/$pid/exe")
+        [[ $executable == "/app/$app" ]]
+        runtime=$(awk '$6 ~ /\/libcoreclr.so$/ {print $6}' "/proc/$pid/maps" | sort -u)
+        [[ $runtime == /app/libcoreclr.so ]]
+        runtime_hash=$(sha256sum "$runtime")
+        printf 'targetPid=%s\nactualExecutable=%s\nmappedCoreClr=%s\ncoreClrSha256=%s\n' \
+            "$pid" "$executable" "$runtime" "${runtime_hash%% *}" > /task/target-runtime.txt
+        observed=1
+        break
+    fi
+    sleep 0.1
+done
+if [[ $observed != 1 ]]; then kill "$vstest_pid" 2>/dev/null || true; fi
+test_status=0
+wait "$vstest_pid" || test_status=$?
+vstest_pid=
+printf '%s\n' "$test_status" > /task/test-exit-status.txt
+if [[ $test_status != 0 ]]; then exit "$test_status"; fi
+[[ $observed == 1 ]]