Re: fis-gtm

[Thorsten Alteholz <debian-med@alteholz.de>]

Hi Andreas,

On Thu, 21 Jul 2011, Andreas Tille wrote:
>         For the packaging itself I wonder if there is really a need for
> asking for a user and group name for the fisgtm user.  I'd say this is
> not common (apache uses www-data, postgresql is using postgres both
> without asking the user for a name.

for whatever reason I had the source of tomcat6 on my computer, had a 
look at it and took it as some kind of template. As fis-gtm should work in 
a clinical environment, there might be some demands on how userids have to 
be named.

>  If you really think that this makes
> sense I would make the question lower priority so that normal users
> will simply go with the default name.

Oops, you are right. I forgot to change it back after some testing.

> Then I tried to build the fis-gtm-initial-i386 package on an amd64
> system (I was sitting in the train to DebCOnf and was offline) and
> perhaps this is the problem - but the install failed (sorry for
> German locale but Thorsten will cope with it)

Yes, that is somewhat ok. In the initial package, there are some binaries 
included that only run on that architecture. So configure must fail.

> - Because of building on an amd64 machine the package ends up with
>   "_amd64.deb" extension.  While I think building on amd64 should
>   not necessarily be forbidden (it's just moving files into a deb)
>   the result should be "_i386.deb"
>   This would probably prevent dpkg from trying to install it on
>   amd64 machines.

Hmm, my original plan was to create two packages, one for i386 and the 
other for amd64. Despite "Architecture: any" in the source control 
file, the Architecture:-line of the binary package control file contains 
i386. So I am afraid that just renaming the package might have some 
unwanted side effects.

> - I also consider the "Architecture: any" in debian/control as not
>   appropriate.  If I'm not missleaded it should rather be i386.

Yes, you are right.

> I have no real clue how to fix this.  It might be an idea if we
> provide both (i386 and amd64) initial tarballs in one source tarball
> and let the rules file detect which one to use depending from the
> architecture the package is builded on.

As upstream provides two tar files, isn't it better to also create two 
packages out of them?

> Kind regards from Banja Luka (just arrived)

I hope the weather is not as bad as here ..

  Thorsten