Re: [PATCH v2 0/7] Add a git regression workflow

[Jeff Layton <jlayton@kernel.org>]

On Wed, 2024-01-03 at 10:08 -0500, Chuck Lever wrote:
> Since the last series, I've added support for tmpfs and NFS mounts,
> in addition to a few other refinements and loads of testing. I'd
> like a little review before I push these.
> 

This all looks pretty great to me. I don't see any need to wait on
merging it, if it seems to be working for you.

> We had talked before about setting up multiple concurrent mounts for
> running tests of different NFS versions in parallel. +1 for running
> tests in parallel, but I think we want to make this a more generic
> mechanism that can test other mount option combinations as well
> multiple NFS versions (and perhaps could be reused with other
> workflows).
>
> I don't have a specific proposal for that yet, but I'm interested in
> discussion.
> 

Version selection is just a mount option anyway. Note that even with
fstests, we now can spin up a separate host for TLS testing, so there is
precedent for non-version host variants anyway.

With fstests now, we can test with these options:

- default mount options
- v4.0
- v3
- TLS (with default mount options)

It probably makes sense to test those configurations for the git
regression suite as well, since they offers pretty good coverage for NFS
client and server.

Are there other configurations that might make sense to test regularly?
It's not too hard to add them.

We probably could even add a "Custom" configuration too that takes
arbitrary mount option and export option strings. That'd make it simpler
to test more exotic configurations.

It would be nice to eventually share this sort of option handling across
different workflows, but that's a more difficult problem given that most
of them handle mounting quite differently.
-- 
Jeff Layton <jlayton@kernel.org>