Bug #42710 Test "mysql_protocols" fails on Solaris 9

Description:
This test failure seems new in cluster-6.4.2, I have never seen it before.

It is directly coupled to Solaris 9:
all runs on Solaris 9, any CPU, any configuration, fail in this way:

======
main.mysql_protocols           [ fail ]

ERROR 2003 (HY000): Can't connect to MySQL server on 'localhost' (128)
mysqltest: At line 9: command "echo "select ' ok' as 'TCP'"       | $MYSQL --protocol=TCP" failed

The result from queries just before the failure was:
<default>
 ok
exec of 'echo "select ' ok' as 'TCP'"       | /PATH/bin/mysql --no-defaults --debug-check --host=localhost  --user=root --password= --port=PORT --socket=/PATH/master.sock --character-sets-dir=/PATH/share/charsets --protocol=TCP' failed, error: 256, status: 1, errno: 29

More results from queries before failure can be found in /PATH/mysql-test/var/log/mysql_protocols.log
======

How to repeat:
Run the tests on Solaris 9.

It *might* be related to bug#37797 (looks somewhat similar),
but I am not sure the different status codes really mean the same thing,
and I doubt a "sporadic" failure on OS X might show up reproducible on Solaris 9.

XRef2: http://tinyurl.com/cwjtqm

I'm not sure whether the URL added yesterday is really helpful:
It leads to a search showing all failures of this test case, regardless what the text is.

When I write "never seen before", I specifically refer to the connect problem whose message I reported.
Adding the terms "localhost" and "protocol" to the search ("test output"), the occurrences are reduced to just "6.0" and "6.0-falcon-team".

Still the same failure in cluster-7.0.5,
still specific to Solaris 9 (all CPUs).

Still present in 7.0.6, without any change.

This also happens on SLES11