[Buildroot] [PATCH] package/tvheadend: bump version to v4.1-2345 and fix version reporting
Baruch Siach
baruch at tkos.co.il
Wed Nov 30 05:34:08 UTC 2016
Hi Arnout,
On Wed, Nov 30, 2016 at 12:28:39AM +0100, Arnout Vandecappelle wrote:
> On 28-11-16 22:03, Thomas Petazzoni wrote:
> > On Mon, 28 Nov 2016 11:32:31 +0100, Damjan Marion wrote:
> >> -TVHEADEND_VERSION = 50a370707aedf5c127e92bb517c378aa8ac04657
> >> +TVHEADEND_VERSION = v4.1-2345-g04ff649
> >
> > Please keep this as a full hash, like it used to be. It should still
> > work fine with the version reporting, if you write this version to
> > $(@D)/rpm/version, like you're doing below.
>
> I'd like to challenge this. We prefer hashes instead of tags because tags may
> still be changed while a hash is guaranteed stable. However, the same can be
> said of tarballs. For tarballs we have the hashes to detect such a situation,
> for VCS downloads we can do the same (if we check hashes).
I don't think we prefer hashes over tags. We do prefer full hashes over
shortened ones. But tags are most preferable, and we have lots of git tags in
<PKG>_VERSION.
> In this particular case, however, it's a github-helper URL so it is a stable
> tarball with a hash, so there is no reason not to use a tag.
>
> Here, it's not actually a tag, but rather than a 'describe' on top of a tag.
> Still, it uniquely identifies a commit, and it's more descriptive than just the
> sha1. For example, with the describe, you can immediately see between two
> versions which one is the most recent one[1]. So I do think there is value in
> providing a 'describe' version rather than just a sha1.
I beg to deffer. In my opinion, v4.1-2345-g04ff649 is misleading because it
gives the impression of a version close to v4.1, which is far from reality.
> There's one issue, however. v4.1-2345-g04ff649 works now because 04ff649 is a
> unique abbreviation. But when later another commit is generated that starts with
> 04ff649, you need to use v4.1-2345-g04ff649a. So I guess it should be
> v4.1-2345-g04ff649a4c13229ebcaa9828fa04339a49842dd5.
That's better than v4.1-2345-g04ff649, but I think that leaving the
'v4.1-2345-g' prefix out is better yet.
baruch
--
http://baruch.siach.name/blog/ ~. .~ Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
- baruch at tkos.co.il - tel: +972.52.368.4656, http://www.tkos.co.il -
More information about the buildroot
mailing list