Re: Ocean Resources also RE: Compiling from scratch

[Ian Jones]

Hi Dennis,

I had similar thoughts with regard to a repository of thorughly tested 
applications and gnustep installation / dependancies. Though if you 
have a certain version of a piece of software tested, it would be 
easier and just as useful to put a url for it and spare yourself the 
bandwidth usage. Same goes for those developers with cvs, maybe if 
people can just give you feedback about what applicationts worked well 
for them on a certain date we could do a cvs co of that code and use 
it, or create tarballs of it if necessary. I'm very keen on getting to 
know cvs at the moment, it seems a pretty wonderful tool and one that 
we could use to our advantage. Maybe application developers would 
consider keeping a semistable tag? Then people could keep reasonable 
current without the risk of being landed with something that doesn't 
work.

I have been working the last 3 days on making gnustep installation 
easier, my work so far is here: 
http://www.digital-bushido.org/opensource/files/GSInstaller.tgz
It still has some problems like you will still need to manually add 
sourcing the GNUstep.sh file to your .bashrc or whatever after the 
install has finished, but for the most part it seems to work. It's a 
little annoying installing to a place that requires root privelages 
though as it configures as a normal user but it will ask you to type 
in the root password if it needs it for you to install for each 
operation as GNUstep modules have to each be built and installed in 
order this seems unavoidable (comments and suggestions welcome on 
this). I have tested it several times today and it seems to 'work for 
me (TM)' but I have no idea how it will behave on other peoples 
systems, so my advice is if you try it, is to just install it to your 
home directory. It will cvs co alexs latest semistable tag, which 
currently only contains core, and put it in ~/GNUstep/gnustep and then 
hopefully configure and install a working system. My normal test has 
been to do the former and then just attempt to compile and install a 
normal application in the same terminal I used to do the install, so 
it uses that same environment and then attempt to run it. I'm not sure 
how people would react to my trying to mess with their .bashrc 
programatically (that's if they use bash). I have learned a lot doing 
this, about subtle differences between fbsd and linux binarys and ones 
that exist on one system and not the another. I had never really done 
a lot with shell scripting before, so it has been quite a learning 
experience all round, I just hope it has been worth the effort and 
that it is of use to someone.

I'm still thinking how to make it easier, and maybe about extending it 
to check properly for dependancies before attempting to do anything. 
Pascal Bourguignon seems keen on the idea of it trying to settle its 
own dependancies, I could try using wget or something for packages 
from a repository such as the one you mention, though I see automated 
installation of system software as pretty risky, but I will look at 
this further at a later date if enough people are interested.

The installer uses two scripts, the only one you will need to run is 
firstStep, the second script is a heavily modified compile-all script 
from core. The compile-all script can still be used independantly and 
hopefully it might replace the one in cvs at some point to save me 
having to copy the file to the core directory once it has been 
downloaded. You could use it standalone for a snapshot tarrball for 
instance. I added some --help so people can have a look what options 
are available. If you are using the firstStep script though you won't 
have to touch configure-all, the choices you make in the firstStep 
script get passed on to the compile-all script and if anything fails 
in the compile-all script it gets relayed back to the firstStep script 
and it will exit.

I still feel I have much to do, but it's a start and hopefully it will 
let people know that I am working at trying to make GNUstep easier to 
install; I'm also very keen to get some feedback and suggestions.

I think I better stop writing now, this was only intended to be a 
quick email :)

Regards,

Ian (Ian_J on #gnustep)

On 2003-10-29 11:33:22 +0000 Dennis Leeuw <dleeuw@made-it.com> wrote:

> Hi all,
>
> I want to proceed with the Ocean project to create a system that 
> works :)
> The initial goal will be to provide a source repository with stuff 
> that is 
> tested on my machine and hopefully works for others too, The sources 
> will be 
> provided at:
> http://ocean.made-it.com/Sources
> The layout is not fixed and will certainly change in the future. But 
> I hope 
> this will already help people with getting started on GNUstep.
>
> Another section I created is:
> http://ocean.made-it.com/Tools
> This currently only contains the start scripts from the Build Guide, 
> but will 
> be expanded with everything that I need to make the system do what I 
> want.
>
>
>
> _______________________________________________
> Discuss-gnustep mailing list
> Discuss-gnustep@gnu.org
> http://mail.gnu.org/mailman/listinfo/discuss-gnustep