selftest history (was: Re: [PATCH] Add libsmbclient list shares test)

Andrew Bartlett abartlet at
Tue Aug 28 23:12:29 UTC 2018

On Tue, 2018-08-28 at 16:02 -0700, Jeremy Allison wrote:
> Yes I already found and fixed that. It still fails
> due to the libsmbclient tests being invoked from
> inside source3/selftest/, whereas Andreas's
> change only fixed the smburl in source4/selftest/
> Whoever thought (a) writing this stuff in python was a good
> idea and (b) naming things identially inside source3 and
> source4 is responsible for a *LOT* of wasted engineering
> time and effort..

What happened is that we had two identical test systems in source3 and
source4.  They were totally independent at one point!

What you are seeing is stuff not merged into common yet (a task clearly

Sadly the definitions are not identical, because the 'samba3' and
'samba4' invocations of the same smbtorture code run with different
options (--target=samba3 or --target=samba4) and produce a different
prefix (samba3. and samba4.), which is encoded in the knownfail files
(et al). 

Finally, to ensure nothing was lost during the merge (stitching might
be a better expression), we run most of our tests twice, many three
times.  A careful eye to removing this duplication would be most

Your assistance to remedy some of this is appreciated.

Andrew Bartlett

Andrew Bartlett
Authentication Developer, Samba Team
Samba Development and Support, Catalyst IT

More information about the samba-technical mailing list