tests: add exit case "KNOWN FAILED"

Now we can mark tests as known to be broken without bothering
users with exit failure. In the build log this "KNOWN FAILED"
may be interpreted as TODO ;)

The main advantage of "known-fail" instead of just skipping
is that we will still find the test diff.

Signed-off-by: Ruediger Meier <ruediger.meier@ga-group.nl>

diff --git a/tests/functions.sh b/tests/functions.sh
index 6631de49d8cea8ad61cdf236f07ef508506e8c84..1dc671d9eb440f2c8c693d7f2291034ee95e0292 100644 (file)
--- a/tests/functions.sh
+++ b/tests/functions.sh
@@ -76,6 +76,15 @@ function ts_failed {
        exit 1
 }
 
+function ts_known_failed {
+       if [ x"$1" == x"" ]; then
+               ts_report " KNOWN FAILED ($TS_NS)"
+       else
+               ts_report " KNOWN FAILED ($1)"
+       fi
+       exit 0
+}
+
 function ts_ok_subtest {
        if [ x"$1" == x"" ]; then
                ts_report " OK"
@@ -196,6 +205,7 @@ function ts_init_env {
 
        TS_VERBOSE=$(ts_has_option "verbose" "$*")
        TS_PARALLEL=$(ts_has_option "parallel" "$*")
+       TS_KNOWN_FAIL=$(ts_has_option "known-fail" "$*")
 
        tmp=$( ts_has_option "memcheck" "$*")
        if [ "$tmp" == "yes" -a -f /usr/bin/valgrind ]; then
@@ -383,6 +393,9 @@ function ts_finalize {
        if [ -s $TS_EXPECTED ]; then
                ts_gen_diff
                if [ $? -eq 1 ]; then
+                       if [ "$TS_KNOWN_FAIL" = "yes" ]; then
+                               ts_known_failed "$1"
+                       fi
                        ts_failed "$1"
                fi
                ts_ok "$1"