Re: [Bacula-beta] regression test changes for root tests

[Dan L. <da...@la...>]

On 6 Sep 2007 at 17:06, Kern Sibbald wrote:

> On Thursday 06 September 2007 16:45, Dan Langille wrote:
> > On 6 Sep 2007 at 9:35, Kern Sibbald wrote:
> > > On Thursday 06 September 2007 03:22, Dan Langille wrote:
> > > > On 5 Sep 2007 at 21:09, Dan Langille wrote:
> > > > > I'm getting regression test failures for root-based tests.  The diff
> > > > > on the restore/original is:
> > > > >
> > > > > # diff tmp/original tmp/restored
> > > > > 50c50
> > > > > < crw--w----  1 1001 4         0   65 2007-09-05 20:09:02 dev/ttyp0
> > > > > ---
> > > > >
> > > > > > crw--w----  1 1001 4         0   65 2007-09-05 20:08:23 dev/ttyp0
> > > > >
> > > > > 84c84
> > > > > < crw-rw-rw-  1 0    0         0   64 2007-09-05 20:09:02 dev/ptyp0
> > > > > ---
> > > > >
> > > > > > crw-rw-rw-  1 0    0         0   64 2007-09-05 20:08:20 dev/ptyp0
> > > > >
> > > > > #
> > > > >
> > > > > I suspect I need to alter this line of tests/dev-test-roo:
> > > > >
> > > > > s%.*[0-9][0-9]:[0-9][0-9]:[0-9][0-9] dev/ttyp1%dev/ttyp1%
> > > > >
> > > > > I guess the script is assuming that we're working on dev/ttyp1.  Why
> > > > > don't we ignore all tty?
> > > >
> > > > FYI, this patch to tests/dev-test-root fixes the problem:
> > > >
> > > > -s%.*[0-9][0-9]:[0-9][0-9]:[0-9][0-9] dev/ttyp1%dev/ttyp1%
> > > > +s%.*[0-9][0-9]:[0-9][0-9]:[0-9][0-9] dev/ttyp%dev/ttyp
> > > >
> > > > Shall I commit it?
> > > >
> > > > On a similar note, I'm finding that the tests/etc-test-root fails
> > > > too:
> > > >
> > > > # diff tmp/original tmp/restored
> > > > 379c379
> > > > < drwxr-xr-x 20 0    0      2560      2007-08-30 09:22:12 etc
> > > > ---
> > > >
> > > > > drwxr-xr-x 20 0    0      2048      2007-08-30 09:22:12 etc
> > > >
> > > > I'm not sure *why* the difference occurs, but I don't think it
> > > > matters.  Shall I patch the test to remove that line from comparison
> > > > too?
> > >
> > > Yes, there are a number of those tests that fail for simple reasons.  It
> > > seems to me that I wrote a special diff that filters most cases.  Are you
> > > using the regular regression script?
> >
> > Do you mean tests/regexwhere-test
> 
> No, I am referring to a program that I think was called testls.  It would do 
> an ls but filtered out certain info that seemed to vary.  I used it in most 
> of the root tests dev-test-root, etc-test-root, lib-tape-root, ...

 === Starting tls-test at 13:56:34 ===

  ===== tls-test OK 13:58:37 ====

That's close!  But perhaps not the same thing.

> That program (testls) is located in the src/tools directory, and if I remember 
> right, its only purpose was to filter certain files ...  Honestly, I forget 
> all the details, because I haven't looked at it for , but you might want to 
> look at it. For about 7 years now.
> 
> The root tests are something that I do not do often, and they are a bit 
> unreliable in that they do often have some false failures due to either the 
> link count or the size field being different -- much as you saw. 

I found many references to testls in tests.  It is being used.

> 
> >
> > > In any case, if you are *sure* that these are false failures (and what I
> > > see does seem to be), it would be a big help if you would filter them.
> >
> > I will.
> 
> OK, thanks.
> 
> 


-- 
Dan Langille - http://www.langille.org/
Available for hire: http://www.freebsddiary.org/dan_langille.php