Re: GNUstep.sh / env sanity patches

[Armando Di Cianno]

On 2004-08-14 02:46:41 -0400 Sheldon Gill <sheldon@westnet.net.au> 
wrote:
> Seemed to me (from quick perusal) that configure.ac changes didn't 
> reflect 
> configure changes.

I have the (gut) feeling that the differences may have been from 
different versions of autoconf ... the important part is in 
GNUmakefile.in anyway.

> Because you asked for it, I've provided an app at the end which 
> contains the 
> essence of your make- and base- patches and hopefully explains things 
> for 
> you.

Thanks. I have the feeling if I work on stuff late at night, I should 
be drinking more coffee. ;-)

> > > I've got GNUstep.sh killed for users but not for developers.
> > > gnustep-make is
> > > too heavily dependent on it at the moment to make it immediately
> > > disappear.
> >
> > Oh, this I would love to help test ... hasn't killing it for users
> > been the only goal?  I suppose developers are uses too. :-) heh.
>
> Not for me it hasn't. As a developer, too, I no longer need to source 
> it at 
> start up. Want to help test? That can be arranged I'm sure...

Yes, absolutely.  For me, this issue is pretty much my only priority, 
atm, in regards to GNUstep, mainly in relation to bringing GNUstep to 
Gentoo, but also the fact that it's stopping people from getting to 
easily use GNUstep.

> I hear and understand your frustration.  I'd like to point out, 
> though, that 
> I did ask you to try and document precisely what build issues were 
> leading 
> you to change $HOME etc so that the *issue* could be identified and 
> properly 
> tested against.

Hrm, yeah, I guess I should have been clearer in my response to this.  
The problem is only "GNUstep packages cause sandbox violations being 
built because GNUmakefiles call tools which try to at least create the 
user Default directory."  So basically, if none of my patches are 
applied, I get "Access violation /root/GNUstep", right from step 1.  
The trouble giving you the exact errors is that as I solve each one, 
piece by piece, new errors arise, but are all of the same form.  So, 
atm, I know that make_services and autogsdoc are big trouble makers, 
as well, as NSUser.m in -base and user_home.c in -make, specifically.

<summary value="important">
Overall, as I may have stated before in previous emails on the lists, 
ideally, the build process should _never_ attempt to write defaults 
(for the package, docs, etc at all) or Defaults should be 
fully/dynamically configurable, this is what I successfully, albeit 
incorrectly wrt memory usage, solved in my last set of patches.
</summary>

So, overally, I'd like to help test this, and move on this, as soon as 
is convenient for you.  If your patches are going to bring this new 
structure to GNustep quickly and efficiently, then let me know how I 
can help test.  This issue is actually causing me sleepless nights, 
which is kind of sad. ;-)

__armando