Re: Request for Sponsorship: VeryFastTree - Parallelized and Optimized Version of FastTree

Étienne Mollier <[email protected]> Mon, 11 Sep 2023 20:11:39 +0200
Newsgroups	gmane.linux.debian.devel.medical
Message-ID	<ZP9YW4nEu3oB7_Dy@fusion>
Hi César,

César Pomar, on 2023-09-11:
> >   * from lintian's automatic review, there are a couple of
> >     warnings about the baseline specific executables caught
> >     without manual pages:
> >
> >         W: veryfasttree: no-manual-page [usr/bin/VeryFastTree-avx2]
> >         W: veryfasttree: no-manual-page [usr/bin/VeryFastTree-avx512f]
> >         W: veryfasttree: no-manual-page [usr/bin/VeryFastTree-avx]
> >         W: veryfasttree: no-manual-page [usr/bin/VeryFastTree-sse2]
> >         W: veryfasttree: no-manual-page [usr/bin/VeryFastTree-sse4_1]
> >
> >     so maybe it would be worth either moving these executables
> >     out of the path, canonically under libexec/ so the dispatch
> >     script finds them there, unless you want to provide the
> >     alternate baseline programs in the path, in which case you
> >     might consider applying a lintian override for those
> >     warnings; or third option, you can modify the manual page to
> >     answer to all the variants;
> 
> Any options seem fine to me, although I don't
> see any scenario where a user would want to run the binaries directly.
> Which option do you recommend?

Given your scenario, I would recommend moving the baseline
specific executables out of the path to libexec/, because it is
not intended that they will be called manually by end users.

The lintian override would be the path of least resistance, but
also a bit boring, and some might consider these binaries might
clutter the Debian binaries namespace.  The third option about
adjusting the manual so the baseline specific programs can be
queried with man is definitely overkill in your context.

> >   * from lintian's pedantic checks, I see also one instance of:
> >
> >         P: veryfasttree source: maintainer-manual-page [debian/VeryFastTree.1]
> >
> >     which makes me wonder whether it would make sense to ship
> >     the manual page straight in your upstream code, for the
> >     benfit of your users not installing veryfasttree through the
> >     Debian package;
> 
> Andreas had already mentioned it to me; the man page is basically the help
> displayed by the binary, so I didn't see the need to distribute it
> outside of Debian.

Acknowledged, oops it looks like I should have had another look
at previous mail exchanges then.  In this case, it may be a good
idea to override the pedantic notice so this doesn't show up on
the radar anymore.

> >   * pristine-tar branch ends with a commit from root: I don't
> >     have strong opinions whether this should be left as-is or
> >     force-pushed with a proper user ID, however I would strongly
> >     suggest not running any packaging step with elevated
> >     privileges in general.
> 
> Sorry, I didn't know that could be considered an issue.
> I ran the build inside a Docker container.

Fair enough…

Please proceed with the two changes I suggested, and the package
should be good for upload.

Cheers,  :)
-- 
  .''`.  Étienne Mollier <[email protected]>
 : :' :  gpg: 8f91 b227 c7d6 f2b1 948c  8236 793c f67e 8f0d 11da
 `. `'   sent from /dev/tty1, please excuse my verbosity
   `-
signature.asc (application/pgp-signature, 833 B)