Lustre devel - Jan 2007 - [Bug 11554] sanity.sh dependencies missing from release tarball

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11554

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|sanity.sh dependencies      |sanity.sh dependencies
                   |missing from release        |missing from release tarball


I think Jim is really referring to what we ship in the tarball.  The 1.4.8
tarball includes the sanity.sh script but not any of the other scripts Jim
refers to.  So aside from the idea of creating a lustre-tests RPM, we need to
fix the tarball.  I have updated the summary accordingly.

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11554



(In reply to comment #3)
> I think Jim is really referring to what we ship in the tarball.  The 1.4.8
> tarball includes the sanity.sh script but not any of the other scripts Jim
> refers to.  So aside from the idea of creating a lustre-tests RPM, we need
to
> fix the tarball.  I have updated the summary accordingly.      
I just noticed myself that b1_5 has the same issue.  sanity.sh is included, but
e.g. oos.sh which it calls is not.

I think probably none of the test dir should be included.  We could include a
simple hardcoded script that starts a local loopback mdt, ost, and client in the
utils directory maybe for a simple "does lustre work" check.

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11554



(In reply to comment #4)

Please don''t remove the tests - we are putting them to good use.

I had hoped that CFS would consider implementing a ''make
check'' target in the
top level Makefile as we have done.  Its good practice and generally easy to run
sanity (and likely other single node tests that we haven''t tried yet)
before
committing changes.

If its not useful to CFS, its at least useful to us.  We generally carry a very
large patch stack, and although we''re pretty good at running IOR and
such on
clusters against candidate release, we''ve not been good about basic
sanity
testing.  Having the ability to detect some regressions with minimal effort
helps.

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11554



Hi Jim,

(In reply to comment #5)
> I had hoped that CFS would consider implementing a ''make
check'' target in the
> top level Makefile as we have done.  Its good practice and generally easy
to run
> sanity (and likely other single node tests that we haven''t tried
yet) before
> committing changes.
Part of our development process (for any patch) involves running
acceptance-small.sh, which includes sanity.  In my case, I build on a different
machine than I test on (my desktop vs. a UML instance) so ''make
check'' would not
be useful.  But I undestand that it would be useful to others, and
wouldn''t
object to seeing one appear at some point :)