On Wednesday 30 March 2005 14:28, Raymond Toy wrote:

> Oops.  That was something that was changed in after 19a.  Previously,
> if :host were not given or given as nil, the host from
> *default-pathname-defaults* was used.  It was changed so explicitly
> giving :host nil would produce an empty host slot so you could merge
> the host from another pathname.

Right. Which is of course an improvement, but I'm still
on 19a.

> There are some restrictions on use :unspecific as the value of a
> pathname component and it's implementation-defined if :unspecific is
> valid.  (See CLHS 19.2.2.2.3.) I guess CMUCL's make-pathname doesn't
> like :unspecific hosts.

Indeed. I wasn't claiming CMUCL's behaviour wasn't
valid. Only that it wasn't what I wanted :-).

>     Gareth> Yay. Oh, but for some reason UFFI doesn't like that.
>     Gareth> It doesn't like #p"cl-library:elephant;libsleepycat.so" either,
>     Gareth> though. So:
>
>     Gareth>            (probe-file (make-pathname :name "libsleepycat"
> :type "so" Gareth>                                       :defaults
> (asdf:component-pathname
>
>     Gareth> (asdf:find-system 'elephant))))
>
> Wouldn't translate-logical-pathname also work here?  Unless you really
> wanted to know if the file existed.

Yes, it would. But I knew about PROBE-FILE and had forgotten
about TRANSLATE-LOGICAL-PATHNAME. Voodoo chicken-waving,
as I said earlier.

Anyway. I think I now (1) understand as much of this as I need to
right now, (2) understand as much as is likely to fit in my brain
unless I devote more time to it than I intend to, and (3) have working
code. So I'll shut up. Thanks for the help!

-- 
g


Reply via email to