Re: OpenMRS package is ready, I believe
Hi Misha,
at first I would really like to thank you for your packaging work on
OpenMRS. If there would be a competition for the fastest learning
person in Debian packaging you would be a quite good candidate. It
is amazing how fast you proceeded from the first steps to a clean
packaging of a non-trivial package. Thanks for your work on this.
On Mon, Oct 04, 2010 at 07:59:25PM -0700, misha680 wrote:
> My apologies for the brief email. I have, I believe, made all changes
> necessary.
If you have made all necessary changes a short mail is fine, isn't it?
;-)
> I tested the package and, to my testing, it works well.
Same for me. Installation leaded to no trouble and the README.Debian
points to the configuration page which works fine. I admit I did not
went through all the configuration pages but I expect it to work or at
least the user has proper advise where to ask for help - that's OK
for me.
Iin your other mail you wrote:
> lintian -I on the changes warns that we are using first person in
> the description. I have checked here
> http://lintian.debian.org/tags/using-first-person-in-description.html
> but we are using "We" in reference to the OpenMRS development team
Well, as I said: lintian -I is "nitpicking mode" and there is no real
need to fix this. So this will not really stop me from uploading and
your ITP bug did not recieved any comments about this from other
developers so you can regard the description as accepted. However, if
you ask me, I think lintian is right here. A description should say
something about the program and not about the mission of the developer
team. You might think about this a bit until the next version is
released - but finally it is *your* package and I will not put my
opinion on your packaging.
I have done a minor change: debian/compat is now 7 instead of 5. The
effect on the resulting package is none, but it makes sense to use
debhelper compatibility level 7 together with debhelper 7. (Probably
you used an old packaging template.)
> Some problems/concerns:
> 1. You may take issue with the pre-dependency on tomcat6. I have
> followed directions here:
> http://fnords.wordpress.com/2010/07/29/how-to-package-your-tomcat-webapp/
> which seemed to work quite well.
>
> If we do _not_ pre-depend on Tomcat, there are some issues:
> * we are installing configuration files directly into Tomcat's /etc/
> folders (semantic issue)
> * more importantly, if we purge _both_ openmrs and tomcat6 at the same
> time, tomcat does not have enough time
> to clear the /var/lib/tomcat6/webapps/openmrs folder _it_ generated,
> and this does _not_ get removed
> (perhaps this is a tomcat6 bug, but due to the semantic issue above I
> believe it is okay to pre-depend here; please
> correct me if I am not mistaken)
When inspecting your former postinst I somehow thought that there might
be reasons to need a predependency. I did not explicitely recommended
it (to not push a predepends in case you might find a better solution)
but I'm not at all astonished that you are coming up with this solution.
> 2. UPDATE: I have added something in postrm purge that removes
> configuration files. This is only configuration for the database and
> modules that are installed - nothing that the user cannot lose.
> However, I still need to test when get home. I have committed changes
> to SVN. If you get a chance to look at other changes at least
> (everything else is tested), much appreciated.
>
> Perhaps a more important technical issue - if we do an aptitude purge
> openmrs that has already been set up,
> the files in /etc/openmrs will _not_ be removed, as they were set up
> by the webapp.
In your other mail you confirmed that purge works (and I also have
tested that the files installed by the openmrs package are removed). I
have not checked whether the databases created by OpenMRS after
configuration will be purged (because I did not went through all the
configuration pages there are probably no more files were created).
> However, to my understanding, purge should remove configuration files
> as well (note: I could do an rm -rf /etc/openmrs in
> postrm, but I noticed this post:
> http://lists.debian.org/debian-devel/2008/06/msg00072.html
> detailing that databases don't remove data on purge? is this true?
I'd consider it the correct purge behaviour if all databases which were
created by the package are purged in the portrm. However, a more
sloppy interpretation is that finally no harm is done if the databases
persist after the purge and the local admin will remove these manually.
So as long as there is no explicite advise in a relevent packaging
document you are more or less free to decide on your own.
I just uploaded the package and tagged the uploaded version in SVN.
Thanks again for your intensive work on this. For Debian Med it is a
real step foreward to get OpenMRS included.
Kind regards
Andreas.
--
http://fam-tille.de
Reply to: