Re: liboqs and oqs-provider

[Roland Illig <roland.illig%gmx.de@localhost>]

Am 10.09.2024 um 23:34 schrieb Jan Schaumann:
> Hey,
>
> It's been I don't know how many years since I've
> created a new package for pkgsrc, so I wanted to
> quickly run this by folks:
>
> I have packages for OpenQuantumSafe's liboqs and
> oqs-provider, the former based on pkgsrc-wip work by
> logan%cyberstorm.mu@localhost.  They install and work for me on
> NetBSD, but I don't know if they follow the current
> best practices etc.
>
> If somebody would like to give it a quick look and see
> if everything looks reasonable, the files are here:
> https://www.netmeister.org/tmp/pqc.tar.gz

Before passing the packages to humans for inspection, you should run
pkglint on them. It reports several stylistic issues and can fix several
of them:

$ pkglint -F security/liboqs security/oqs-provider

When you then run:

$ pkglint -Wall security/liboqs security/oqs-provider

you get some warnings.

I'm not sure how to fix the warning about TOOLS_PATH.python3. There
might be a more elegant way of making a python3 program available at
test time, but I don't know about it.

You don't need to set BUILDLINK_ABI_DEPENDS at all, just declaring the
API dependency suffices.

You can ignore the warning about the unknown shell command "ctest".
Pkglint should find out by itself that including devel/cmake/build.mk
makes this command available.

In oqs-provider/Makefile, the trailing semicolons after the mkdir
commands are redundant, but they don't hurt.

Everything else looks fine to me.

Roland