On Fr, 2013-10-04 at 13:40 +0200, Dominik George wrote:
Vincent Lefevre vinc...@vinc17.net schrieb:
Then, do you mean that VCS hashes are sortable?
Of course not. One would have to do something like 0~MMDDnn
+git in that rare case.
My argument for keeping the VCS hash is to ease
Thanks !
On 15/10/13 22:32, Benjamin Drung wrote:
On Fr, 2013-10-04 at 13:40 +0200, Dominik George wrote:
Vincent Lefevre vinc...@vinc17.net schrieb:
Then, do you mean that VCS hashes are sortable?
Of course not. One would have to do something like 0~MMDDnn
+git in that rare case.
On 2013-10-02 16:51:09 +0200, Dominik George wrote:
Dominique Dumont d...@debian.org schrieb:
well, you proposed a version like 'hg'. if upstream switches to
git, you can't use a version like 'git' because it sorts before
hg. I grant you that is easy to work around.
If you deem it
On 2013-10-02 17:50:40 +0200, Dominik George wrote:
I established an advantage for the user using my proposal - go get
me a disadvantage for the packager.
As a user, I dislike long version strings.
That said, what's the point in NOT being verbose?
Version strings need to be displayed, and if
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512
Vincent Lefevre vinc...@vinc17.net schrieb:
Then, do you mean that VCS hashes are sortable?
Of course not. One would have to do something like 0~MMDDnn+git in that
rare case.
My argument for keeping the VCS hash is to ease identifying
On 2013-10-04 13:40:29 +0200, Dominik George wrote:
My argument for keeping the VCS hash is to ease identifying the code
in the package.
Does it need to be in the version string?
Why not somewhere else?
The goal of the Version field in Debian packages is to identify
and sort several versions
Hello,
On 04/10/13 13:50, Vincent Lefevre wrote:
On 2013-10-04 13:40:29 +0200, Dominik George wrote:
My argument for keeping the VCS hash is to ease identifying the code
in the package.
Does it need to be in the version string?
Why not somewhere else?
The goal of the Version field in
On Fri, Oct 04, 2013 at 01:31:35PM +0200, Vincent Lefevre wrote:
On 2013-10-02 17:50:40 +0200, Dominik George wrote:
That said, what's the point in NOT being verbose?
Version strings need to be displayed, and if they take too much space,
they may be truncated (e.g. in aptitude).
They can
Hello,
I am packaging a versionless library software maintained via a mercurial
repository.
Is there any custom for this case ?
If not, can we use the version format 'hgMMDD' ?
Best regards,
Jerome
--
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512
Jerome BENOIT g62993...@rezozer.net schrieb:
Hello,
I am packaging a versionless library software maintained via a
mercurial repository.
Is there any custom for this case ?
If not, can we use the version format 'hgMMDD' ?
Best regards,
On Wednesday 02 October 2013 15:21:48 Jerome BENOIT wrote:
I am packaging a versionless library software maintained via a mercurial
repository. Is there any custom for this case ?
If not, can we use the version format 'hgMMDD' ?
As a user, I don't care about upstream repo.
You should use
Hi Nik,
thanks for your quick reply.
On 02/10/13 15:49, Dominik George wrote:
Jerome BENOIT g62993...@rezozer.net schrieb:
Hello,
I am packaging a versionless library software maintained via a
mercurial repository.
Is there any custom for this case ?
If not, can we use the version
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512
Hi,
What does 'XX' stand for ?
The short commit hash, as proposed in your initial mail.
You should use a version derived from the date only. This way, you
won't be in
trouble if upstream switches to git.
I absolutely do not see why this
Hello,
On 02/10/13 16:05, Dominik George wrote:
Hi,
What does 'XX' stand for ?
The short commit hash, as proposed in your initial mail.
In my first email, what you read as a commit hash was meant to be the date.
To summarise:
0~MMDD
should be fine.
Thanks,
Jerome
You
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512
0~MMDD
should be fine.
It isn't, it is not a unique identifier for the one release you are packaging.
- -nik
-BEGIN PGP SIGNATURE-
Version: APG v1.0.8-fdroid
iQFNBAEBCgA3BQJSTCxtMBxEb21pbmlrIEdlb3JnZSAobW9iaWxlIGtleSkgPG5p
+++ Dominik George [2013-10-02 15:49 +0200]:
Jerome BENOIT g62993...@rezozer.net schrieb:
Hello,
I am packaging a versionless library software maintained via a
mercurial repository.
Is there any custom for this case ?
I tend to use:
0~MMDD+hgXX
It sorts just below anything
+++ Dominik George [2013-10-02 16:23 +0200]:
0~MMDD
should be fine.
It isn't, it is not a unique identifier for the one release you are
packaging.
No, but it can be a sufficient identifier so long as you don't make more
than one release a day.
Which exact tag/branch/hash/whatever
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512
Wookey woo...@wookware.org schrieb:
+++ Dominik George [2013-10-02 16:23 +0200]:
0~MMDD
should be fine.
It isn't, it is not a unique identifier for the one release you are
packaging.
No, but it can be a sufficient identifier so long as
On Wednesday 02 October 2013 16:05:18 Dominik George wrote:
You should use a version derived from the date only. This way, you
won't be in
trouble if upstream switches to git.
I absolutely do not see why this should be an issue.
well, you proposed a version like 'hg'. if upstream
Wookey wrote:
+++ Dominik George [2013-10-02 15:49 +0200]:
Jerome BENOIT g62993...@rezozer.net schrieb:
Hello,
I am packaging a versionless library software maintained via a
mercurial repository.
Is there any custom for this case ?
I tend to use:
0~MMDD+hgXX
It sorts just
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512
Dominique Dumont d...@debian.org schrieb:
On Wednesday 02 October 2013 16:05:18 Dominik George wrote:
You should use a version derived from the date only. This way, you
won't be in
trouble if upstream switches to git.
I absolutely do not see
Wookey woo...@wookware.org (2013-10-02):
The 'use an ISO date as version' idea comes from advice in the
developer packaging docs somewhere. It would be good if this 0~ trick
was mentioned there too so one could decide whether to use it or not
at the time of initial packaging.
On Wednesday 02 October 2013 16:51:09 Dominik George wrote:
well, you proposed a version like 'hg'. if upstream switches to
git, you
can't use a version like 'git' because it sorts before hg. I grant
you
that is easy to work around.
If you deem it unlikely that two commits are made
Hi,
On 2 October 2013 17:27, Dominique Dumont d...@debian.org wrote:
If you deem it unlikely that two commits are made in the same day (which
happens all the time), how likely is it that upstream switches VCSs and
does an important commit on the same day?
that's not the issue. Try that:
On Wednesday 02 October 2013 17:31:02 Andrew Shadura wrote:
dpkg --compare-versions 1.hg2012 '=' 1.git2013 || echo 'false'
Weren't we talking about 0~20131002.hg2efc4fcd vs 0~20131002.git67ed491a?
Sorry, I confused between Jerome original mail and Dominik's proposal.
Dominik's idea raises
Quoting Dominik George (2013-10-02 16:39:09)
Wookey woo...@wookware.org schrieb:
+++ Dominik George [2013-10-02 16:23 +0200]:
0~MMDD
should be fine.
It isn't, it is not a unique identifier for the one release you
are packaging.
No, but it can be a sufficient identifier so long as you
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512
A packager is not required to serve users with such specific needs.
Hmm, I last saw that attitude when being explained the Arch way.
I established an advantage for the user using my proposal - go get me a
disadvantage for the packager.
That
Hi,
Wookey woo...@wookware.org (2013-10-02):
It would be good if this 0~ trick
was mentioned there too
Already in the New Maintainers' Guide:
http://www.debian.org/doc/manuals/maint-guide/first#namever
Regards
David, kinda ashamed to add a message to that longish nitpicking thread
Jerome BENOIT g62993...@rezozer.net writes:
I am packaging a versionless library software maintained via a
mercurial repository. Is there any custom for this case ?
I have had a surprising rate of success simply asking upstream to make
versioned release tarballs, or at least VCS tags for
Hello All,
I did not expected to initial such a thread: thanks a lot for all the messages
On 02/10/13 20:24, David Prévot wrote:
Hi,
Wookey woo...@wookware.org (2013-10-02):
It would be good if this 0~ trick
was mentioned there too
Already in the New Maintainers' Guide:
30 matches
Mail list logo
