[aur-general] TU Application - Dustin Falgout

Levente Polyak anthraxx at archlinux.org
Wed Mar 2 21:48:27 UTC 2016

On 03/02/2016 09:19 PM, Balló György wrote:
> If you don't specify tag or commit hash at the end of the git source, then
> you should use the -git suffix. Users expect if the package has no -git
> suffix, then it's a working static version tested by the maintainer, and
> not some experimental code from git HEAD.

Its not just a should but a guideline rule [0] that must be followed
upon. For official repositories it is mandatory and will also be
enforced because of multiple reasons: the most obvious ones are rebuilds
on sobumps and reproducible packages (not yet there but a topic that is
being worked on).
The only difference is that (besides that the AUR is unsupported) on the
AUR people may not notice it or care enough to enforce that. However, in
my personal opinion, a trusted user should do things above the general
average, that's IMHO why someone should be _trusted_.

On 03/02/2016 08:21 PM, Dustin Falgout wrote:
> I do have a sane reason indeed. Upstream is not following their
> github releases. If you look in openSUSE's package repo you will see >
that they are packaging the latest master as the most recently
> released version. Looking at the history of those packages it seems
> that whoever is maintaining the packages over at openSUSE does not
> use github releases in their release process on a regular basis.

It applies for just one out of 3 packages, you should fully check your
claim before using it as an argument. Also 2/3 of those packages are not
something that could be considered "not released on regular basis"

- last release: Sep 3, 2015
- patch commits since release: 4
- openSUSE version [1]: 0.5.3 release

- last release: Jun 1, 2015
- patch commits since release: 9
- openSUSE version [2]: 0.5.3 0b4ce51 (2 patch commits since release)

- last release: Nov 5, 2013
- patch commits since release: 7
- openSUSE version [3]: 7897d3f (7 patch commits since release)

Actually, as mentioned above, its just the recompress packages that
really falls out of scope. The tar_scm is just a debian control file fix
and a missing extension parameter to service file.
I don't see any point why the release is not sufficient for those. Also
did you try to contact upstream about a recompress release?

On 03/02/2016 08:21 PM, Dustin Falgout wrote:
> Considering that, it doesnt make sense to tag the end of the pkgname >
with "-git"

As mentioned in my very first section, this part is not something that
can be argued upon [0], there are just two sane possibilities:
1) use a static version and have the pure package name
2) use non-static VCS (like git HEAD) and add such postfix to pkgname


[0] https://wiki.archlinux.org/index.php/VCS_package_guidelines#Guidelines

More information about the aur-general mailing list