failing tests in the testsuite on debian

[Jeremy Allison]

On Tue, Nov 15, 2022 at 04:35:45PM +0300, Michael Tokarev via samba-technical wrote:
>I experimented with the tests some more. Thank you Jeremy for the
>tip about escaping spaces in the knownfail files, - somehow I noticed
>this is the case in other places but didnt' think about using escapes
>in my "knownfails".
>
>But the thing, or a request for help actually, is still here.  I found
>out that samba testsuite is unreliable.
>
>In particular, one can't re-run tests, because on subsequent runs, even
>if you `rm -rf st' in-between, the result will be different, and different
>tests will fail and the ones failed before will succeed, even with the
>--quick test list.
>
>eg, samba3.rpc.schannel_anon_setpw anonymous password set (schannel enforced server-side)(nt4_dc_schannel)
>test will succeed on the first run, but will fail on subsequent runs:
>
> Failed with dcerpc_flags=0x800220
> WARNING!: ../../source4/torture/rpc/schannel.c:796: status was 
>NT_STATUS_CONNECTION_DISCONNECTED, expected NT_STATUS_OK: Failed to 
>connect without schannel
> Failed with dcerpc_flags=0x800220

That's not supposed to happen. I personally always do

rm -rf st

before re-running tests and doing so should always
cause re-creation of the servers.

>On the other hand, samba4.rpc.echo against rpc proxy with domain creds(rpc_proxy)
>fails on the first run (as described in my previous email), but succeeds
>on subsequent runs.
>
>I'm asking for help to make the whole thing runnable as a "quick" check
>that the new samba works fine and there's no obvious build errors.

I'll take a look at all tests that fail unexpectedly
after a rm -rf st, but it's a lower priority than
code reviews and feature fixes.