Libguestfs - May 2022 - [libguestfs PATCH 4/4] tests/regressions: remove "iface"-based restrictions

On Thu, May 05, 2022 at 08:25:35AM +0200, Laszlo Ersek
wrote:
> On 05/04/22 20:50, Richard W.M. Jones wrote:
> > 
> > For the whole series:
> > 
> > Reviewed-by: Richard W.M. Jones <rjones at redhat.com>
> > 
> > Do you think it makes sense to completely drop the two regression
> > tests in patch 4?  They're testing something that (probably)
cannot go
> > wrong now.
> 
> We can do that, but then a few more test cases should be modified
> perhaps, for consistency. If you grep the tree for "iface",
you'll find
> a few test cases that pass various "iface" params (off-hand I
recall
> Ruby, OCaml, and Java-language test cases -- I think those are all
> "bindtests"). I figured those tests still made sense (for
ensuring API
> compat), so I didn't remove the iface params, and/or the test cases
that
> only differrerd from the rest in their iface params. And then I figured
> the rhbz regression tests should stay too (just be relaxed).
We cannot drop the parameter, it's part of the ABI.
> If we drop the latter (the rhbz regression tests), should we still keep
> the other (language bindings) tests?
The bindtests are mainly about testing the binding between language
and the C library so without actually looking IIRC 'iface' is only
used to check that strings are converted correctly.

Anyway, it's fine to leave it.

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
nbdkit - Flexible, fast NBD server with plugins
https://gitlab.com/nbdkit/nbdkit

On 05/05/22 10:48, Richard W.M. Jones wrote:
> On Thu, May 05, 2022 at 08:25:35AM +0200, Laszlo Ersek wrote:
>> On 05/04/22 20:50, Richard W.M. Jones wrote:
>>>
>>> For the whole series:
>>>
>>> Reviewed-by: Richard W.M. Jones <rjones at redhat.com>
>>>
>>> Do you think it makes sense to completely drop the two regression
>>> tests in patch 4?  They're testing something that (probably)
cannot go
>>> wrong now.
>>
>> We can do that, but then a few more test cases should be modified
>> perhaps, for consistency. If you grep the tree for "iface",
you'll find
>> a few test cases that pass various "iface" params (off-hand I
recall
>> Ruby, OCaml, and Java-language test cases -- I think those are all
>> "bindtests"). I figured those tests still made sense (for
ensuring API
>> compat), so I didn't remove the iface params, and/or the test cases
that
>> only differrerd from the rest in their iface params. And then I figured
>> the rhbz regression tests should stay too (just be relaxed).
> 
> We cannot drop the parameter, it's part of the ABI.
I didn't mean dropping the parameter; I meant dropping the testing of
the parameter in the bind tests.

However:
>> If we drop the latter (the rhbz regression tests), should we still keep
>> the other (language bindings) tests?
> 
> The bindtests are mainly about testing the binding between language
> and the C library so without actually looking IIRC 'iface' is only
> used to check that strings are converted correctly.
You are right that as long as the parameter is part of the API
(regardless of whether it is used for anything useful, or just ignored),
we should continue testing *its binding*.
> Anyway, it's fine to leave it.
Commit range 4864d21cb8eb..ddf276884c04.

Thank you!
Laszlo