On (28/05/15 17:53), Pavel Reichl wrote: > > >On 05/28/2015 05:45 PM, Lukas Slebodnik wrote: >>On (28/05/15 17:25), Pavel Reichl wrote: >>>Hello, >>> >>>please see simple attached patch. Although test catches the bug and is no >>>doubt useful I believe that we should avoid segfaulting even in tests. >>> >>>Thanks for considering! >>>From a509c49fbc55664170d96e0e29bf6263c8d38a2a Mon Sep 17 00:00:00 2001 >>>From: Pavel Reichl <[email protected]> >>>Date: Thu, 28 May 2015 11:13:47 -0400 >>>Subject: [PATCH] TESTS: fix segfault in test_sss_strerror_err_last >>> >>>If there were more error codes than error messages. This >>>test segfaulted. >>> >>When test crashed then it failed. >I don't think that relying on undefined behavior is a good practice. It's not undefined behaviour. becuase you should not dereference NULL pointer.
>>So the problem was not in test but in sssd code. >Well technically the test is part of SSSD code suite IMO. Test can crash if the bug is in the code. result of make check is non zero -> failed. >>IMHO it's sufficient to test that the last error code is the same. >IMHO it's not good if test is based on occurrence of segfault. I do not agree with exporting global variables just for test. It's very BAD practice. If it was not clear then: _ _ _ ____ _ __ | \ | | / \ / ___| |/ / | \| | / _ \| | | ' / | |\ |/ ___ \ |___| . \ |_| \_/_/ \_\____|_|\_\ LS _______________________________________________ sssd-devel mailing list [email protected] https://lists.fedorahosted.org/mailman/listinfo/sssd-devel
