Re: [Sbcl-devel] SBCL 1.0.9 doesn't build on NetBSD

[Aleksej S. <as...@ho...>]

Christophe Rhodes <cs...@ca...> writes:

> Aleksej Saushev <as...@ho...> writes:
>
>> Thanks for the hint, I'll suggest the patch (see attachment) to
>> pkgsrc maintainers. With this patch I'm successful to build SBCL
>> 1.0.8 under pkgsrc framework (that is: update version number,
>> add patch, "make distinfo", "make").
>>
>> For 1.0.9 I get fail with attached log. Any suggestions?
>
> Build with a lisp that is not CLISP.  (This is precisely the kind of
> non-repeatable build corruption that I referred to in my last message
> -- non-repeatable in that trivial changes make the problem go away.)

There's no other Common Lisp implementaion.
Can you suggest specific one, which could be compiled within usual
UNIX and alike framework?

>>> For this and other reasons -- notably the non-repeatable build
>>> corruption that I have seen off and on for years using clisp -- I
>>> wouldn't recommend using it to bootstrap sbcls.
>>
>> There's no other portable (and free!) Common Lisp implementation
>> to use for bootstrapping, CLISP just works. Sometimes.
>>
>>> I appreciate the theoretical purity for an OS distributor to be able
>>> to bootstrap sbcl without having a lisp compiler already, using clisp,
>>> but I think that if you want to do that you need to spend some time
>>> diagnosing clisp's particular intermittent difficulties with the sbcl
>>> build and genesis process.
>>
>> pkgsrc is very nice framework, it's stable and works in any sane
>> environment.  And I don't see the reason, why anyone needing, say, 
>> Maxima, should fight installation of SBCL, which should be done in
>> rather unconvenient way.
>
> Is it not the case that pkgsrc could use binary versions of SBCL to
> build SBCL, rather than depending on whatever random clisp version is
> currently in the system?  How does gcc get bootstrapped?

GCC is not bootstrapped, you can have no GCC at all,
and pkgsrc specifically supports native C compilers,
where it is appropriate.