From: phillip.wood123@gmail.com
Date: Wed, 28 Aug 2024 14:58:50 GMT
Subject: Re: [PATCH v6 00/13] Introduce clar testing framework
Message-ID: <13c94c27-0975-46dc-b94e-72fd3972ef16@gmail.com>
In-Reply-To: <Zs8uDnUAFFKF4DTF@tanuki>

Hi Patrick

On 28/08/2024 15:03, Patrick Steinhardt wrote:
> On Wed, Aug 28, 2024 at 02:18:49PM +0100, Phillip Wood wrote:
>> On 20/08/2024 15:02, Patrick Steinhardt wrote:
>>> Hi,
>>>
>>> this is another version of my patch series that introduces the clar
>>> testing framework for our C unit tests.
>>
>> I've left some comments on the code, most of them are points I made in the
>> last round that received no response. My main concern is that the assertions
>> offered by clar are not as convinent as the check_* macros.
> 
> Did you have a look at my reply at <ZsSTOczuGhT5zwoV@tanuki>, where I
> responded to these concerns?

Oh sorry I'd completely missed that mail (it came in while I was off 
line and I failed to notice it).

> In summary: I'm aware that this is still a rough edge. I'd be happy to
> follow up on this and improve usability of the assertions, but doing it
> likely is a bit more involved, mostly because I want to upstream all
> changes in this context. So I'd rather want to land a basic version
> first, and then I'd iterate and improve asserts.

That sounds reasonable

>> What's the plan for converting our current tests if this gets merged? If we
>> were to add wrappers that provide check_int() etc. that would greatly
>> simplify the conversion. I think it would offer a more ergonomic api for
>> writing new tests than the verbose and non-typesafe cl_assert_equal_i() and
>> friends.
> 
> My plan would first be to let things cook for a bit while I sort out the
> rough spots upstream. Once done and once we are sufficiently sure that
> this is the direction to go I'm happy to do the conversion myself.
> 
> Whether we want to have wrappers... I dunno, I don't think the names are
> all that bad. They have a clear namespace and say rather directly what
> they are doing, which I value more than briefness (to a certain extent,
> of course).

One could argue the check_* are namespaced by "check". I find writing 
unit tests in C is pretty tedious and having to type cl_assert_equal_? 
just adds to that.

> The type safety is another topic though, and something I
> will aim to address.

Thanks, it sounds like we're more or less on the same page

Best Wishes

Phillip

> Thanks!
> 
> Patrick