knownfail and subunit errors

[Andrew Bartlett]

On Thu, 2012-04-26 at 07:59 +1000, Andrew Bartlett wrote:
> On Wed, 2012-04-25 at 11:44 +0200, Andrew Bartlett wrote:
> 
> > commit e346bba1fc369cc13438cf3882961ef7b11ccc3a
> > Author: Andrew Bartlett <abartlet at samba.org>
> > Date:   Wed Apr 25 16:26:50 2012 +1000
> > 
> >     s4-torture: enable raw.composite test again, and convert to new style test
> >     
> >     This test has been skipped for some time, but is an important test for parallel operation.
> >     
> >     Andrew Bartlett
> > 
> > -----------------------------------------------------------------------
> > 
> > Summary of changes:
> >  selftest/knownfail                           |    1 +
> >  selftest/skip                                |    2 -
> >  source4/lib/messaging/messaging.c            |    3 +-
> >  source4/libcli/smb_composite/fsinfo.c        |   13 +-
> >  source4/libcli/smb_composite/smb_composite.h |    6 +-
> >  source4/torture/raw/composite.c              |  233 ++++++++++++++------------
> >  source4/torture/raw/raw.c                    |    2 +-
> >  7 files changed, 143 insertions(+), 117 deletions(-)
> > 
> > 
> > Changeset truncated at 500 lines:
> > 
> > diff --git a/selftest/knownfail b/selftest/knownfail
> > index 959e3d2..82ff0d6 100644
> > --- a/selftest/knownfail
> > +++ b/selftest/knownfail
> > @@ -35,6 +35,7 @@
> >  ^samba3.blackbox.rpcclient over ncacn_np with \[spnego,smb2,bigendian\]
> >  ^samba3.blackbox.rpcclient over ncacn_np with \[spnego,connect,smb2\]
> >  ^samba3.blackbox.rpcclient over ncacn_np with \[spnego,connect,smb2,bigendian\]
> > +^samba3.raw.composite .appendacl # This needs ACL and IDMAP behaviour that s3 cannot provide, including mapping of arbitary sids.
> >  # these show that we still have some differences between our system
> >  # with our internal iconv because it passes except when we bypass our
> >  # internal iconv modules
> 
> Jelmer,
> 
> I'm confused about the acceptable syntax for knownfail, as this entry
> works on autobuild, the sn-devel-104 build farm host and my development
> workstation (Fedora 16), but fails on these other build farm hosts:
> 
> https://build.samba.org/build.cgi/build/c47e78a98af9770c99e13f9b1701b2504e855c5d#lnk-test-566 and https://build.samba.org/build.cgi/build/925edd267394e01f50cc4209e0d51a9c147b970b#lnk-test-565
> 
> What am I doing wrong with this knownfail entry?  It looks like many of
> the others in the file.

For those playing along at home, this was fixed by:

https://gitweb.samba.org/?p=samba.git;a=commitdiff;h=ed9f903b12e6c38e0d467a7b0722fa0410056cf2

That is, when migrating a test to the proper subunit/torture_assert()
code, it must be done completely or else things like this happen.

Andrew Bartlett
-- 
Andrew Bartlett                                http://samba.org/~abartlet/
Authentication Developer, Samba Team           http://samba.org