Re: Correction to web site information: OpenMRS has debian packages available for download
On Tue, Sep 21, 2010 at 09:29:15AM -0700, misha680 wrote:
> Thank you all so much for your wonderful comments.
> I have made an initial attempt at a "proper" debian folder:
> http://old.nabble.com/file/p29771184/debian.zip debian.zip
> Still very early stages, although it does compile for OpenMRS 1.6.1:
Thanks for your honest work on this. I had a look into your files and I
would like to suggest moving it to the Debian Med repository to enable
other Debian Med members easy access to the code. To enable you commit
permissions you should register at alioth.debian.org and once you have
a guest login just ask for becoming a member of the Debian Med team.
You can find my changes to your work at:
1. debian/watch file is somehow not working - we have to sort out this
2. Cosmetical change: I renamed your debian/orig-tar.sh to
debian/get-orig-source. In principle there is absolutely nothing
wrong with the old name - we just have this convention in Debian Med
and I added an according target in debian/rules
3. My basic changes are in (this renamed) debian/get-orig-source.
I added several comments. Please look into the diff carefully and
ask if there are questions remaining.
IMPORTANT CHANGE: I stripped the .svn directories because it is
heavily recommended in upstream tarballs. This change actually
leads to a build problem later and I wonder why this happened.
IMHO a build should NOT relay on existing .svn directories so I
would rather recommend fixing the build process than keeping the
.svn directories. But that's an upstream issue.
Meta-Comment in this direction: I would strongly recommend issuing
release tarballs instead of asking people to check out SVN. This
would not only make our life (namely uscan) easier it is IMHO more
convenient for users
4. debian/changelog: A common missunderstanding about this file is
that you should list upstream changes here but the correct use is
to mention only packaging changes. For instance in case you might
switch from sun-java-jdk to open-jdk or something like this should
be mentioned here. In our initial packaging state we just mention
"Initial packaging" because logically before there any official
package exists there can not be any changes to the previous packaging.
At least this is ftpmasters point of view - but we can agree to
list some packaging changes from the can use it for the moment for
our internal communication. Example entries would be
* renamed debian/orig-tar.sh to debian/get-orig-source
- obtain version number if not given at command line
- parse source package name from changelog
- remove .svn dirs (attention: breaks build!)
* use Debian Med list as maintainer for group maintenance
* put myself and Misha as Uploaders
* added VCS-fields
* updated Standards-Version
* s/Build-Depends-Indep: sun-java6-jdk/Build-Depends-Indep: openjdk-6-jdk/
this might or might not break the build / running program - just
trying if we really need Sun java
I commited the changes to the SVN above and hope you like it. Feel free
to revert any change in case you do not agree (*you* are the maintainer
and have the final word - my role is just mentoring and finally sponsoring
your work but not overriding any of your decisions).
> I had a few questions.
> 1. lenny or squeeze?
> is written based on squeeze.
You technically found the answer and used "unstable" in the changelog.
That correctly answers your question. Remark: It is a *very* good idea
to use an unstable pbuilder environment. I'm pretty sure you can setup
such a Debian unstable build environment on your Ubuntu machine and so
your are safe from both sides: You actually build against Debian and you
are using the correct distribution target. (Feel free to ask if you
have no idea about what I'm talking about.)
> I tried to package on lenny, but did not have latest versions. I tried
> nightly build of squeeze but it failed to install in VMWare (I checked 3x).
> Now am downloading weekly build to try.
Just try pbuilder which is a good idea anyway instead of fiddling around
with virtual machines.
> The debian folder I attached is based on simply using my Ubuntu system.
I really guess the behaviour of
will be pretty much the same on Ubuntu - it would be nice if you can
sort out the missing .svn dir issue.
> 2. This I believe is _important_.
> Thank you for the helpful example of an svn package:
> unfortunately, the reference SVN server in the watch file
> which is
> uses <a href tags. Our server
> uses <file href tags.
> uscan does _not_ process these correctly:
> Any ideas much appreciated.
Well, probably waiting until this bug is fixed (or helping yourself
fixing the bug which I personally would not spent my time currently).
I have the feeling that at the point in time whe have sorted out all
other problems which might occure in OpenMRS packaging this problem
might have been fixed without any action from our side. Alternatively
you might consider releasing OpenMRS in release tarballs which would
circumvent this problem easily. Currently we do not really need a
working uscan because the get-orig-source target just does the job we
want it to do.
Hope this helps