Re: [PATCH v2 0/3] mergesort: move tests to Clar and retire the helper

From: Patrick Steinhardt <ps@pks.im>
Date: Oct 9, 2026, 11:52 UTC
Message-ID: <asjVkwUYJI6wERWf@pks.im>
In-Reply-To: <cover.1791365181.git.dilsheddilu123@gmail.com>

Hi,

On Wed, Oct 07, 2026 at 07:20:22PM +0530, Muhammed Dilshad A wrote:

Show 5 quoted lines

> Hi Patrick,
> 
> Thanks for the review. I followed up on the larger cleanup you mentioned.
> The sorting tests now run in Clar, and I have removed the old benchmark
> and its helper. This also removes the unused generate subcommand.

please reply to reviews individually instead of replying in the cover letter.

> The new suite keeps all 1,680 cases from the old certification test and
> adds checks for empty and small lists using both sort macros. It checks
> sorting order, stability and list length. Cleanup frees the backing
> arrays directly, so a failed assertion does not need to walk list links.

It would have made it easier to review if the new tests were added in a separate commit.

Show 10 quoted lines

> Changes since v1:
> 
> * Patch 1 is unchanged.
> * Patch 2 moves the tests to Clar and removes the unused generate and
>   test commands. The sort command remains available for the benchmark.
> * Patch 3 removes p0071 and the remaining sort helper, along with their
>   build and command registrations.
> 
> I kept the leak fix first so it can still be applied on its own if you
> would prefer to keep the broader cleanup for a separate series.

I dunno, I feel like that's not quite useful. If we didn't want to take the broader cleanup we'd instead apply v1 of your seires. In this version of the patch series it's plain unnecessary churn because we remove the code anyway.

> The Make and Meson unit tests pass, and the mergesort unit suite also
> passes with LeakSanitizer enabled. The production sorting implementation
> is unchanged.

This information is quite curious, as it makes me wonder why it is even noteworthy to point out. My basic assumption is that folks who send a series to the mailing list test their stuff, so there is no need to explicitly say so.

I mean I of course know why this is here: it's the typical "let's check all the boxes" output that AI is so happy to generate. *sigh*

Patrick