On 9/9/19 9:35 AM, Trevor Gamblin wrote:
From: Trevor Gamblin
The libevent ptest used to report only a global pass or a fail result.
Count individual PASS, FAIL, SKIP results. The SKIP results now
include tests that are disabled in the libevent code.
libevent's ptest output did not comply with the automake-style output
"result: testname", and reported a FAIL status at the end of the test
run if any of the libevent tests failed. This patch makes the log
consistent with the automake style:
PASS: http/cancel_by_host_no_ns
PASS: http/cancel_inactive_server
PASS: http/cancel_by_host_no_ns_inactive_server
SKIP: http/cancel_by_host_server_timeout
SKIP: http/cancel_server_timeout
and provides a summary as follows:
=== Test Summary ===
TOTAL: 316
PASSED: 300
FAILED: 0
SKIPPED: 16
DURATION: 87
END: /usr/lib/libevent/ptest
Signed-off-by: Trevor Gamblin
---
.../libevent/libevent/run-ptest | 43 ---
.../libevent/libevent_2.1.11.bb | 3 ++
2 files changed, 31 insertions(+), 15 deletions(-)
diff --git a/meta/recipes-support/libevent/libevent/run-ptest
b/meta/recipes-support/libevent/libevent/run-ptest
index 0241851c70..79d6da3ee1 100644
--- a/meta/recipes-support/libevent/libevent/run-ptest
+++ b/meta/recipes-support/libevent/libevent/run-ptest
@@ -1,18 +1,31 @@
#!/bin/sh
-fail=0
-for test in ./test/*
-do
- $test
- if [ $? -ne 0 ]
- then
- fail=1
- fi
-done
+# run-ptest - 'ptest' test infrastructure shell script that
+# wraps the libevent test scripts
+#
+# Trevor Gamblin
+###
+LIBEVENTLIB=@libdir@/libevent
+LOG="${LIBEVENTLIB}/ptest/libevent_ptest_$(date +%Y%m%d-%H%M%S).log"
-if [ $fail -eq 0 ]
-then
- echo "PASS: libevent"
-else
- echo "FAIL: libevent"
-fi
+cd ${LIBEVENTLIB}/ptest
+
+# Run the libevent regress test, but format the pass/fail outputs so
+# that ptest can count them. Ignore the line containing "tests or
+# "TESTS" as that line is the libevent total at the end of the test
+# and is counted as an extra test failure, e.g.
+# "2/300 TESTS FAILED. (31 skipped)" will be seen as a test failure
+# rather than a total
+./test/regress 2>&1| sed -e '/TESTS/d' -e '/tests/d' -e '/OK/ s/^/PASS: / ;
/FAILED/ s/^/FAIL: / ; /SKIPPED/ s/^/SKIP: / ; /DISABLED/ s/^/SKIP: /' | cut -f1,2 -d
':' | tee -a ${LOG}
+
+passed=`grep PASS ${LOG}|wc -l`
+failed=`grep FAIL ${LOG}|wc -l`
+skipped=`grep -E SKIP ${LOG}|wc -l`
+all=$((passed + failed + skipped))
+
+( echo "=== Test Summary ==="
+echo "TOTAL: ${all}"
+echo "PASSED: ${passed}"
+echo "FAILED: ${failed}"
+echo "SKIPPED: ${skipped}"
+) | tee -a ${LOG}
diff --git a/meta/recipes-support/libevent/libevent_2.1.11.bb
b/meta/recipes-support/libevent/libevent_2.1.11.bb
index 1e18f0ab2c..f005ab8bda 100644
--- a/meta/recipes-support/libevent/libevent_2.1.11.bb
+++ b/meta/recipes-support/libevent/libevent_2.1.11.bb
@@ -43,4 +43,7 @@ do_install_ptest() {
do
install -m 0755 $file ${D}${PTEST_PATH}/test
done
+
+# handle multilib
+sed -i s:@libdir@:${libdir}:g ${D}${PTEST_PATH}/run-ptest
}
Noticed that the autobuilder results using v2 were detecting libevent's
"2/300 TESTS FAILED. (31 skipped)" line at the end of the regress test
as an extra failure. Fixed that up and reduced the ptest to just running
"regress", as the other scripts in the test folder are more
performance-oriented and don't provide a pass/fail result that can be used.
--
___
Openembedded-core mailing list
Openembedded-core@lists.openembedded.org
http://lists.openembedded.org/mailman/listinfo/openembedded-core
