BIND 10 #606: See which of BIND 9 tests can be re-used or re-implemented for BIND 10

[BIND 10 Development]

#606: See which of BIND 9 tests can be re-used or re-implemented for BIND 10
-------------------------------------+-------------------------------------
                 Reporter:  stephen  |                Owner:  jinmei
                     Type:  task     |               Status:  reviewing
                 Priority:  major    |            Milestone:  A-Team-
                Component:           |  Sprint-20110309
  b10-auth                           |           Resolution:
                 Keywords:           |            Sensitive:  0
Estimated Number of Hours:  3.0      |  Add Hours to Ticket:  0
                Billable?:  1        |          Total Hours:  0
                Internal?:  0        |
-------------------------------------+-------------------------------------

Comment (by jinmei):

 Replying to [comment:10 vorner]:

 > > To be honest I didn't think about that much; it was basically a
 > > straightforward port of the equivalent BIND 9 code (written in C).
 > > On thinking about it, I see some subtle difference: with explicit
 unlink,
 > > we can make sure that a file is created even if there's an existing
 one
 > > for which the current process doesn't have the write persmission (as
 long
 > > as the directory is writable).
 >
 > On the other hand, if the directory is not writable, but the file is
 (prepared for the process), the simple version works, while this one
 doesn't.

 When the directory is not writable we can't even create the PID file
 in the normal scenario (note that bind10 cleans it up on normal
 shutdown), so the "simple" version won't be useless most of the time.

 That said...

 > > But this is relatievely a minor difference.  If you want to remove the
 > > explicit unlink, I'm okay with that.
 >
 > I think for the sake of simplicity, we might want to remove it.

 ...at least at the moment the PID file is only intended to be used for
 tests, where the above failure cases won't matter much anyway.  So I
 accepted your preferred way, and removed the unlink.

 > > >  - The last line of the test is kind of confusing. It starts with E:
 which probably means End, but I thought it was error for a while. I had to
 examine the exit code to see it terminated successfully.
 > >
 > > Hmm, I agree 'E' is confusing.  But I'd keep it for now to be
 > > compatible with BIND 9's framework as much as possible (e.g. there may
 > > be a post-processing script that assumes this notation and we may want
 > > to reuse it without modifying it).  For that matter, some other
 > > notations are not clear to me (I don't know what "A" or "I" actually
 > > means, for example).  I think we should clarify (as a separate task)
 > > these with BIND 9 developers, and if agreed, update the both notations
 > > in a consistent manner.
 >
 > Or add a documentation about what they mean somewhere, at last.

 I'll handle this as a separate ticket as I don't even know what some
 of them mean.

 > Anyway, the make distcheck still fails for me :-(. This time when it
 tries to distclean:

 Okay, I was lazy:-) and sorry for bothering you again.

 This time I made it sure by performing distcheck and creating and
 using a tar ball.  While I did it I noticed other necessary files are
 missing in the tar ball even if they are not used in distcheck, and I
 fixed this by adding everything to EXTRA_DIST.

 I also noticed some other points that were not covered in the
 discussion so far, and fixed them:

 - changed test_dump_pid_notexist.  It was not what I actually intended
   to test and what it did didn't make sense (although it didn't fail).
 - in system tests use port 53210 instead of 5300 to be aligned with #523
 - remove the copy of digcomp.pl and testsock.pl and refer to these in
   the BIND 9 source tree (this system test already relies on it).
   these files don't have to be modified for BIND 10 specific purposes,
   so sharing the same file would make more sense.

-- 
Ticket URL: <http://bind10.isc.org/ticket/606#comment:12>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development