Re: [Postfixadmin-devel] xmlrpc.php, tests/* and some possible bugs

[Christian B. <pos...@cb...>]

Hello,

Am Samstag, 11. April 2009 schrieb Christian Boltz:
> [several possible bugs]

I commited the fixes as discussed on IRC.

When I now run tests/run.php, I _sometimes_ get 1 failure.

Postfixadmin XMLRPC Unit Tests
1)  at [/home/cb/postfixadmin/HEAD/tests/RemoteAliasTest.php line 44]
        in testUpdateRemoteOnly
        in RemoteAliasTest
        in ./RemoteAliasTest.php
FAILURES!!!
Test cases run: 4/4, Passes: 26, Failures: 1, Exceptions: 0

Some testing shows that it might be a race condition.
Adding a sleep(1) to RemoteAliasTest.php (see patch below) fixes the 
problem and the testcases work without a failure.

Do you have any idea what could cause the race condition in this
specific test?

--- RemoteAliasTest.php (Revision 626)
+++ RemoteAliasTest.php (Arbeitskopie)
@@ -41,7 +41,7 @@
     public function testUpdateRemoteOnly() {
         $this->assertTrue($this->alias->update(array('ro...@ra...'), 'remote_only'));
         $this->assertFalse($this->alias->hasStoreAndForward());
-        $this->assertTrue($this->alias->update(array('ro...@ra...'), 'remote_only'));
+        sleep(1); $this->assertTrue($this->alias->update(array('ro...@ra...'), 'remote_only'));
         $this->assertTrue($this->alias->update(array('ro...@ra...', 'fi...@fi...', 'ro...@ru...'), 'remote_only'));
         $this->assertEqual($this->alias->get(), array('ro...@ra...', 'fi...@fi...', 'ro...@ru...'));
         $this->assertFalse($this->alias->hasStoreAndForward());


Regarding error_reporting:
> IMHO run.php should ini_set() it to be really sure we catch all
> undefined variables etc. which won't cause noticable errors ("just"
> strange bugs) otherwise.

Unfortunately setting error_reporting in tests/run.php doesn't help much.
The reason is that the tests do http calls to xmlrpc.php, and therefore the
setting in php.ini is used on the xmlrpc.php side...

Another small issue: The tests should clean up the database after being 
run IMHO. (Or at least use a random password for the mailbox created by the
tests to avoid abuse if someone runs the tests on a public server.)


Regards,

Christian Boltz
-- 
Das ist die Goldene Regel für das Performancetuning von
UNIX-Systemen: RAM ist nur durch mehr RAM zu ersetzen.
[Kristian Koehntopp in suse-linux]