Re: cannot open heap

On Tue, 30 Jul 2002, Serguei Panskikh wrote:

> Gary Byers said:
> >
> >
> > On Mon, 29 Jul 2002, Serguei Panskikh wrote:
> >
> >> Hi,

[a lot of stuff about not being able to load an image.]

> the image is not there. The last successful step was the build of binary
> code of the kernel. Than I've tried to build the image using commands above
> and it did not work. I might use wrong commands to build the image but that
> was the only about kernel invocation I found in the documentation.

There seems to be a misunderstanding here.

You need a lisp image (as provided in the binary distribution) to compile
the lisp sources.  Once you have compiled those sources into FASL (.pfsl)
files, you can build a lisp image.   The process of building a lisp involves
loading a bunch of FASL files into a "bootstrapping image" and saving
the result.  The process of creating a "bootstrapping image" involves
compiling and loading some code from the "level-0" directory.  All of
this is described in the "lisp-build.html" document.

Once you've built an image this way, you can repeat the process as
often as you like and discard any binaries that you haven't produced
from source yourself.

It is generally no easier to compile lisp sources without a lisp compiler
than it is to compile C sources without a C compiler.

I was about to suggest that you read the "lisp-build.html" document
more carefully; I did so myself, and realized that that document doesn't
explicitly state that one needs to use a (binary) version of the lisp
to compile the lisp and produce one's own copy.  Depending on one's
point of view, that seems almost too obvious to mention; different people
might have different points of view, so it's almost certainly worth
stating explicitly.

If you understand all of this and have downloaded the binary distribution,
you might want to ensure that it's installed in the same "ccl" directory
as the sources are.

>
> Once again, neither ./ppccl nor ./openmcl work being called without
> arguments from appropriated directories, they return the same error message:
> "Couldn't load lisp heap image from ./PPCCL"
>

Without arguments (and without an image) this is exactly what they should
say.  You'd indicated earlier that you always got a message referencing
"./PPCCL" regardless of what arguments they were called with; it's hard
to understand how that could have happenned.

>
> >
> >>
> >> I guess I've followed the way described in the avaliable
> >> documentation, but ... I understand that I may have done something
> >> wrong. Or my OS (YDL 2.1) might be wrong. But as I followed the
> >> documented procedure then it's certainly a bug of documentation. Or of
> >> distribution - there is no ./configure file, is it?
> >>
> >
> > There isn't a "configure" file: OpenMCL currently runs on one
> > architecture and two platforms, and (so far) there isn't a whole lot to
> > configure.
>
> Good, then the problem is not with my system.
>
> >
> >> Anyway, how to build it? I've checked a binary distro of OpenMCL - it
> >> works (it runs), but it doesn't help me - we have a policy - every
> >> open source software must be build from sources before using it. Isn't
> >> it open source?
> >>
> >
> > Everything distributed with OpenMCL can be built from the sources
> > provided.
>
> Very good. Then I'll try to succeed with you help, right?

If that policy's carried to the extreme, you wouldn't want to use the
C compiler unless it had been built from source.  Since the C compiler's
written in C, you'd need a C compiler to build it.  Since that C compiler's
written in C ...

>
> P.S. One authour of another open source project answered me: "Why would you
> want ot build it from sources? Just download pre-build binaries and enjoy!"
> That doesn't work. But that might be offtopic here.
>
> >
> > I believe that the procedure for doing so is fairly simple and fairly
> > clearly documented, though I suppose that I'm not an objective judge of
> > either of those issues.  I have no motivation for intentionally making
> > the process complicated or making the documentation obscure, and if I'm
> > doing either or both of those things unintentionally I'd certainly want
> > to know.
>
> I believe you. I just mean that if it is not a bug in my system, then it
> might be either bug in software or in documentation or I do something wrong.

From what I can tell, it (now) appears that you expected to be able to
build the lisp without using the binary copy that's provided.  That can't
work.

>
> --
> Best Regards,
> Serguei Panskikh
>
>
>
> -----------------------------------------
>    "Webmail by SquirrelMail!"
> http://squirrelmail.org/
>
>
>
>