[PATCH v3 2/4] support: Add FUSE-based file system test framework to support/

[DJ Delorie]

Florian Weimer <fweimer@redhat.com> writes:
> The addresses &f->prepared and &f->prepared.attr are the same because
> it's a union.

Ok.

>>> +  struct fuse_out_header outh =
>>> +    {
>>> +      .len = sizeof (outh) + payload_size,
>>> +      .unique = f->inh->unique,
>>> +    };
>>
>> outh.error not set?
>
> It's set to zero due to the initialization.  I think that's okay?

I had it in mind that "local variables without initializers are not
initialized" and applied that to local structure fields without
initializers.  My bad.

>> Ok.  Is there any need to ensure the second thread exited?  I assume it
>> will exit when the main thread exits.
>
> It doesn't exist due to the support_fuse_cleanup_pid != getpid () check.
> This code is running before the fork.  It doesn't work with --direct,
> but I don't quite see how to handle the cleanup reliably in that case,
> sorry.

Ok.

>> Do we want to use xstatx (instead of statx) if the whole point is to
>> test the interface?  Or is "it failed" in xstatx good enough?
>
> It's more about testing the FUSE thread, so I think xstatx is fine.

Ok.