Re: Bsdtar and archive torture tests

Ed,

Have you done anything with this?  I'm very interested in
getting some real regression tests imported and this seems
like a great place to start.

Ideally, the test would be run several times:

    * archive with gnutar/extract with bsdtar
    * archive with cpio/extract with bsdtar
    * archive with bsdtar/extract with cpio
    * archive with bsdtar/extract with gnutar
    * etc...

Unfortunately, the exact tests here will vary slightly:

    * gnutar can archive sparse files that bsdtar should
      correctly extract, but neither bsdtar nor cpio can
      archive sparse files

    * bsdtar can archive (and restore) very large directories
      and very deep directory trees.  My testing of this has
      been somewhat hampered by limits in the rest of the
      system: /bin/sh won't cd to a dir whose path is longer
      than about 8k, rm/find/ls/du are all limited to 32k
      path lengths, etc.

I suspect the ideal test arrangement would provide switches
to the comparison routine to omit/ignore certain files.
Then you could build a single "original", make several copies
using the above combinations, then compare while overlooking
any unsupported attributes.

Thoughts?

Tim

Ed Maste wrote:
> On Wed, Sep 28, 2005 at 05:43:09PM -0400, Kris Kennaway wrote:
> 
> 
>>Can we import the tar stress tests as a regression test?  i.e. what is
>>the license on them?
> 
> 
> Right now the only mention of any copyright info is
> # Copyright 2003, Elizabeth D. Zwicky
> in one of the files.
> 
> The author states
>     The test programs were written in a spirit of experimentation,
>     rather than with the intention of producing software for other
>     people to use.  I strongly encourage people who are interested
>     in testing backup and archive programs to produce their own
>     tests that cover the cases they are most interested in.  However,
>     if you insist on using my programs, or just want to snicker at
>     my programming, they are available from
>     http://www.greatcircle.com/~zwicky
> 
> Despite that, I think they've already demonstrated their value.
> 
> The stress test consists of two perl programs.  One creates a test tree,
> and the other compares file metadata between the original tree and a
> restored tree.  An automated method for running the test and comparing
> the results is not included (but would be easy to write, of course).
> 
> I'll contact the author to ask about the license.
> 
> --
> Ed Maste, Sandvine Incorporated
> _______________________________________________
> freebsd-current_at_freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
> 
>