Re: New package managment

[Ingo Ruhnke <grumbel@gmx.de>]

Steve Baker <sjbaker1@airmail.net> writes:

> (Only because of all the example programs and their data - it would
> only be about 250Kb without all that stuff)

Didn't knew that, since I had never had a look at the examples before,
but hey really nice :-)  
 
> ...and a better name. Erm, how about "autoload" with the file extension
> being ".al"?

how about autoget? But lets wait for a name if we have something.
  
>   eg: mygame.al would contain...
               ^^
Better leaving it to .sh, just more standard and instead call it
something like: mygame-version-autoget.sh
 
>    if ( -e /usr/local/plib/lib/libssg.so.1.0 ) then
>      wget http://www.woodsoup.org/projs/plib/plib.al
>      ./plib.al
>    endif

The check is just to simple, it would fail if I installed the library
for example to '/usr/local/lib' or '/opt' or what ever.

> Obviously it needs to be more sophisticated than that - it's got to
> check the version number of the library that's installed and the
> acceptable range of library versions for this package - also
> there would need to be LOTS of error checking and the ability to
> go and look at mirror sites for the packages it grabs.

All right, but for the libabry checking it would be more error save to
use a mini configure script, since autoconf allready has all this
error checking build in.

> I was assuming that you would need to provide a configuration file
> for the actual autoload program that would generate the mygame.al
> script.  I don't thing there is the right kind of information present
> in any of the existing autoconf/automake stuff.

If the configure script is modulary build, the checks could be done
with the macros in acinclude.m4, but I never tested if that is
practical. 
 
> In fact, you could simply delegate the test for library presence
> to the library.al script.  That would mean redundantly downloading
> lots of library.al scripts for libraries that are already installed,
> but those scripts should be tiny - and the advantage would be to
> punt the issue of how to detect the presence of that library in
> the correct version range onto the author of the library - which
> is probably A GOOD THING.

Yep, sounds like a good idea, that would also minimize the possibility
of pure/buggy library checks.

-- 
                                  http://dark.x.dtu.dk/~grumbel/pingus/ | 
Ingo Ruhnke <grumbel@gmx.de>             http://home.pages.de/~grumbel/ |
------------------------------------------------------------------------+