Index | Thread | Search

From:
Stuart Henderson <stu@spacehopper.org>
Subject:
Re: help with porting libaec
To:
Marco van Hulten <marco@hulten.org>
Cc:
ports@openbsd.org
Date:
Thu, 3 Jul 2025 13:43:57 +0100

Download raw body.

Thread
On 2025/07/03 12:23, Marco van Hulten wrote:
> There are also a libaec tarballs available with a generated configure
> script, but the one from the main mirror has a URL that changes
> unpredictable with new versions, as shown here:
> https://gitlab.dkrz.de/k202009/libaec/-/issues/6#note_344239

the naming of release artefact directories is a gitlab misfeature, imho.
using the release name would be much simpler..

from your issue linked above,

: To this end, I am using two SITES (for redundancy), namely
: 
:     https://gitlab.dkrz.de/k202009/libaec/-/archive/
:     https://github.com/MathisRosenhauer/libaec/releases/download/

I recommend not doing that.

I suggest just using the github url with the predictable dir name,
there's no general problem with using the gitlab one and updating the
dir as needed, but it's more hassle for updates.

> Are there reasons (except for success in this case) for using a tarball
> either with or without a configure script?

I prefer the ones with, then you know that it was generated with a
version of autoconf (and general environment) known to the software's
author. One less thing to think about if there are build problems.

> CMake is apparently the preferred method for libaec, but is there any
> advantage there?

Not really as long as the other method isn't rotten upstream, imho.