On Mon Oct 09 20:39:45 2017, pemensik@redhat.com wrote:
> Hello,
> 
> just a small note for tsiggss test on Fedora 26 (and later as well). It
> will pass just fine if started with
> $ KRB5_CONFIG=/dev/null sh run.sh tsiggss
> 
> There is KRB5_CONFIG environment in tests.sh right at the start. However
> named server is started by start.pl script before prep and tests.sh are
> started. It inherits system defaults. If KRB5_CONFIG=/dev/null is
> exported from conf.sh, test will pass again. I am surprised it works on
> Fedora 25 without a change, but I think it is test configuration issue.

Thanks for the debugging on Fedora, this is much appreciated.

Here's the patch I pushed to our repo:

diff --git a/bin/tests/system/conf.sh.in b/bin/tests/system/conf.sh.in
index 0d63f129e8..3672e1c662 100644
--- a/bin/tests/system/conf.sh.in
+++ b/bin/tests/system/conf.sh.in
@@ -65,6 +65,9 @@ MAKEJOURNAL=$TOP/bin/tests/makejournal
 PIPEQUERIES=$TOP/bin/tests/system/pipelined/pipequeries
 SAMPLEUPDATE=$TOP/lib/samples/sample-update

+# we don't want a KRB5_CONFIG setting breaking the tests
+KRB5_CONFIG=/dev/null
+
 # The "stress" test is not run by default since it creates enough
 # load on the machine to make it unusable to other users.
 # v6synth
@@ -150,6 +153,7 @@ export KEYFRLAB
 export KEYGEN
 export KEYSETTOOL
 export KEYSIGNER
+export KRB5_CONFIG
 export MAKEJOURNAL
 export MDIG
 export NAMED
diff --git a/bin/tests/system/tsiggss/tests.sh b/bin/tests/system/tsiggss/tests.sh
index 42d1db0dd8..2fb850717f 100644
--- a/bin/tests/system/tsiggss/tests.sh
+++ b/bin/tests/system/tsiggss/tests.sh
@@ -15,10 +15,6 @@ status=0

 DIGOPTS="@10.53.0.1 -p 5300"

-# we don't want a KRB5_CONFIG setting breaking the tests
-KRB5_CONFIG=/dev/null
-export KRB5_CONFIG
-
 test_update() {
     host="$1"
     type="$2"

Ondrej