> >>>>> "rif" == rif  <[EMAIL PROTECTED]> writes:
> 
>     rif> Create-target works fine, I go into setenv.lisp and uncomment only the
>     rif> "specific features that most people want" and "version tags"
>     rif> sectionss.  build-world.sh seems to be OK, then rebuild-lisp.sh gives
>     rif> many complaints about 
> 
> Don't touch any of the setenv.lisp settings.  You shouldn't have to do
> that if you just to rebuild the current version with the same set of
> features.

<snipped> 

> That it doesn't for you means that we've probably not been clear enough.
> 
> Ray
> 

First off, let me say that I really appreciate the help that everyone
has been giving me.  I hope what follows will be perceived as useful
suggestions for improving the docs, rather than criticisms, and I hope
that as my knowledge about the system increases, I can provide
something useful in return to everyone who's been helping.

I've now built a basic core using the 04-2004 sources and 04-2004
binary.  In terms of clarity, as far as I know, the only documentation
to look at is the README file in /src/tools/.  It might be worthwhile
explicitly mentioning there that linux users want to use the
linux_gencgc rather than the linux option (I don't even know what this
means); this is not clear a priori.  There is also a paragraph in the
README (lines 200-205) which states:

  IMPORTANT: You will normally have to modify the sample setenv.lisp
  file, to include code that puts at least a minimal set of features
  onto the list (use PUSHNEW and/or REMOVE).  You can use the current
  set of *features* of your lisp as a first guide.  The sample
  setenv.lisp includes a set of features that should work for
  Linux/x86.

This paragraph makes it seem that maybe you're OK as it stands on
Linux, but if you go look at the setenv.lisp you get after running
create-target.sh, you see towards the top a line "Specific features
that most people want:", and then a bunch of things commented out.  To
me, the combined effect of the IMPORTANT paragraph and the setup of
setenv.lisp and the "step-by-step example" (where things care cat'ted
into a setenv.lisp) make it seem natural to start uncommenting things
in setenv.lisp --- my guess was that I would basically have no
features if I didn't uncomment anything.  Now it seems that the "base
list" of features is the features of the host CL executable you're
using to compile?  Actually, looking closer at the docs, this is
already stated, but it's only done where it talks about
cross-compiling (lines 284-290), I'd probably also mention it
somewhere at the top.

Now that I have a lisp.core, and I can say

./linux/lisp/lisp -core ./linux/lisp/lisp.core -noinit -nositeinit

What do I have to do to finish (i.e., make this functionally
equivalent to the CMUCL binary I started with)?  If I try taking away
the -noinit or -nositeinit, I get crashes.  AFAICT, they're happening
when my init file tries to load clocc and is barfing on something
gray-streams related:

;    - Loading binary file "/home/rif/Software/clocc/src/port/gray.x86f"
;; Loading #p"/home/rif/Software/clocc/src/port/gray.x86f".
 
MK:DEFSYSTEM: missing system GRAY-STREAMS.
   [Condition of type MAKE:MISSING-SYSTEM]
 
Restarts:
  0: [CONTINUE] Return NIL from load of "/home/rif/Software/clocc/src/port/gray.x86f".
  1:            Return NIL from load of "home:.cmucl-init".
  2: [ABORT   ] Skip remaining initializations.

How do I "finish"?

Cheers,

rif

Reply via email to