Re: [Sbcl-devel] hppa-hpux 1.0.24.8

[Larry V. <re...@us...>]

> On Sat, Jan 3, 2009 at 11:18 PM, Larry Valkama
> <re...@us...> wrote:
> 
>> Remade the stub so it is only for hpux. hppa-linux shouldn't need it.
>> The confusion in chenegc.c and breakpoint.c is removed, wasn't needed. I
>> thought calculation of code would get wrong but testing without it on
>> hardware confirms it works without.
>>
>> The 1-vop is a mix of the old hppa-code and the mips backend.
> 
> Thanks! Everything should now (as of 1.0.24.22) be merged -- given
> that, how would you characterize the status of the HPUX port?
> 
> I added a couple of apparently missing #+hpuxs into the 0-stub2 patch:
> I hope they were right...
> 
> Additionally, that patch contained the following:
> 
> +#ifdef LISP_FEATURE_HPUX
> +    return_from_lisp_stub = (void *) ((char *)*((unsigned long *)
> +                 ((char *)initial_function - 1)) + 23);
> +#endif
> 
> I merged this as well, but I would strongly prefer a if the magic
> numbers here could be replaced with named constants or defines.
> 
> Cheers,
> 
>  -- Nikodemus
> 
Thanks alot for the commit!

Small fix in src/compiler/hppa/system.lisp needed:

 #!+hpux  <-- was #+hpux
(define-vop (setup-return-from-lisp-stub)

What I've done with it so far:
Load swank and slime can connect remotely.
Install, load and tried cl-ppcre using asdf/asdf-install.
Havn't dared to cd tests yet.

There is still stuff left to be done. For example SIMPLE-ERROR will
sigsegv and retreat to repl instead of invoking debugger.
Also (DEBUG some-high-number) will sigsegv.
Am sure tracing and ldb wont work.

cheers