bugfix and testbench improvements

- bugfix: solved potential (temporary) stall of messages when the queue was
  almost empty and few new data added (caused testbench to sometimes hang!)
- fixed some race condition in testbench
- added more elaborate diagnostics to parts of the testbench
- solved a potential race inside the queue engine

6 files changed, 37 insertions, 1 deletions

diff --git a/tests/arrayqueue.sh b/tests/arrayqueue.sh
index 01fc133b..58fd24ae 100755
--- a/tests/arrayqueue.sh
+++ b/tests/arrayqueue.sh
@@ -10,5 +10,8 @@ source $srcdir/diag.sh injectmsg  0 40000
 
 # terminate *now* (don't wait for queue to drain!)
 kill `cat rsyslog.pid`
+
+# now wait until rsyslog.pid is gone (and the process finished)
+source $srcdir/diag.sh wait-shutdown 
 source $srcdir/diag.sh seq-check 0 39999
 source $srcdir/diag.sh exit
diff --git a/tests/da-mainmsg-q.sh b/tests/da-mainmsg-q.sh
index 6ec2f3a9..d502fca3 100755
--- a/tests/da-mainmsg-q.sh
+++ b/tests/da-mainmsg-q.sh
@@ -7,7 +7,7 @@
 # check everything recovers from DA mode correctly.
 # added 2009-04-22 by Rgerhards
 # This file is part of the rsyslog project, released  under GPLv3
-echo "testing main message queue in DA mode (going to disk)"
+echo "[da-mainmsg-q.sh]: testing main message queue in DA mode (going to disk)"
 source $srcdir/diag.sh init
 source $srcdir/diag.sh startup da-mainmsg-q.conf
 
diff --git a/tests/daqueue-persist-drvr.sh b/tests/daqueue-persist-drvr.sh
index 7b6ec6dd..d95991fc 100755
--- a/tests/daqueue-persist-drvr.sh
+++ b/tests/daqueue-persist-drvr.sh
@@ -26,5 +26,6 @@ source $srcdir/diag.sh check-mainq-spool
 echo "#" > work-delay.conf
 source $srcdir/diag.sh startup queue-persist.conf
 source $srcdir/diag.sh shutdown-when-empty # shut down rsyslogd when done processing messages
+$srcdir/diag.sh wait-shutdown
 source $srcdir/diag.sh seq-check 0 4999
 source $srcdir/diag.sh exit
diff --git a/tests/diag.sh b/tests/diag.sh
index 56810c4f..d8ba43b8 100755
--- a/tests/diag.sh
+++ b/tests/diag.sh
@@ -43,6 +43,7 @@ case $1 in
 		then
 		   echo "ABORT! core file exists, starting interactive shell"
 		   bash
+		   exit 1
 		fi
 		;;
    'wait-queueempty') # wait for main message queue to be empty
diff --git a/tests/linkedlistqueue.sh b/tests/linkedlistqueue.sh
index 9570ed2b..72c2a403 100755
--- a/tests/linkedlistqueue.sh
+++ b/tests/linkedlistqueue.sh
@@ -10,5 +10,8 @@ source $srcdir/diag.sh injectmsg  0 40000
 
 # terminate *now* (don't wait for queue to drain)
 kill `cat rsyslog.pid`
+
+# now wait until rsyslog.pid is gone (and the process finished)
+source $srcdir/diag.sh wait-shutdown 
 source $srcdir/diag.sh seq-check 0 39999
 source $srcdir/diag.sh exit
diff --git a/tests/nettester.c b/tests/nettester.c
index 71641745..2838b919 100644
--- a/tests/nettester.c
+++ b/tests/nettester.c
@@ -47,6 +47,7 @@
 #include <signal.h>
 #include <netinet/in.h>
 #include <getopt.h>
+#include <errno.h>
 
 #define EXIT_FAILURE 1
 #define INVALID_SOCKET -1
@@ -90,6 +91,7 @@ void readLine(int fd, char *ln)
 	if(verbose)
 		fprintf(stderr, "begin readLine\n");
 	lenRead = read(fd, &c, 1);
+
 	while(lenRead == 1 && c != '\n') {
 		if(c == '\0') {
 			*ln = c;
@@ -102,6 +104,11 @@ void readLine(int fd, char *ln)
 	}
 	*ln = '\0';
 
+	if(lenRead < 0) {
+		printf("read from rsyslogd returned with error '%s' - aborting test\n", strerror(errno));
+		exit(1);
+	}
+
 	if(verbose)
 		fprintf(stderr, "end readLine, val read '%s'\n", orig);
 }
@@ -308,6 +315,10 @@ processTestFile(int fd, char *pszFileName)
 
 		/* pull response from server and then check if it meets our expectation */
 		readLine(fd, buf);
+		if(strlen(buf) == 0) {
+			printf("something went wrong - read a zero-length string from rsyslogd");
+			exit(1);
+		}
 		if(strcmp(expected, buf)) {
 			++iFailed;
 			printf("\nExpected Response:\n'%s'\nActual Response:\n'%s'\n",
@@ -372,11 +383,24 @@ doTests(int fd, char *files)
 	return(iFailed);
 }
 
+
+/* indicate that our child has died (where it is not permitted to!).
+ */
+void childDied(__attribute__((unused)) int sig)
+{
+	printf("ERROR: child died unexpectedly (maybe a segfault?)!\n");
+	exit(1);
+}
+
+
 /* cleanup */
 void doAtExit(void)
 {
 	int status;
 
+	/* disarm died-child handler */
+	signal(SIGCHLD, SIG_IGN);
+
 	if(rsyslogdPid != 0) {
 		kill(rsyslogdPid, SIGTERM);
 		waitpid(rsyslogdPid, &status, 0);	/* wait until instance terminates */
@@ -457,6 +481,9 @@ int main(int argc, char *argv[])
 	}
 	fclose(fp);
 
+	/* arm died-child handler */
+	signal(SIGCHLD, childDied);
+
 	/* start to be tested rsyslogd */
 	openPipe(testSuite, &rsyslogdPid, &fd);
 	readLine(fd, buf);
@@ -467,5 +494,6 @@ int main(int argc, char *argv[])
 		ret = 1;
 
 	if(verbose) printf("End of nettester run (%d).\n", ret);
+
 	exit(ret);
 }