[asdf-devel] MCL issue

[Chun Tian (binghe)]

Hi, Faré

I think the "binding ccl::*user-homedir-pathname*" suggestion by Terje was based on "not touching ASDF source code itself". With this approach, people could use

(let ((ccl::*user-homedir-pathname*
       (ccl::findfolder #$kuserdomain #$kCurrentUserFolderType)))
  (asdf:load-system ...))

to make ASDF work on MCL.

For your question about why not the conditional be inside the truename, i think it's because MCL's result are already a truename:

? (in-package :asdf)
#<Package "ASDF">
? (ccl::findfolder #$kuserdomain #$kCurrentUserFolderType)
#P"Macintosh HD:Users:binghe:"
? (pathname-directory-pathname *)
#P"Macintosh HD:Users:binghe:"
? (truenamize *)
#P"Macintosh HD:Users:binghe:"

MCL doesn't support UNIX-like pathname (i.e. "/Users/binghe"), so above is just the final truename.

I'm not sure if it's good to consider using a MCL patch to change the return value of (user-homedir-pathname), because Common Lisp standard never say what exactly the "user's home directory" means when the optional host arguments not given [1]. I guess MCL designers may thought that MCL application itself is living in a separate "host", apart from the Macintosh OS (as another host), this is a common approach in old Lisp systems like Symbolics Lisp Machine as I can see.

(when I'm still writing this mail, Terje already showed another viewpoint: the patch may conflict with legacy code)

Any way, as a MCL & ASDF user, I suggest using my previous patch, and I don't disagree following form:

(defun* user-homedir ()
  (truenamize (pathname-directory-pathname #-mcl (user-homedir-pathname)
                                           #+mcl (ccl::findfolder #$kuserdomain #$kCurrentUserFolderType))))

Regards,

Chun Tian (binghe)

[1] http://www.lispworks.com/documentation/HyperSpec/Body/f_user_h.htm#user-homedir-pathname

在 2011-5-19，10:58， Faré 写道：

> 2011/5/18 Chun Tian (binghe) <binghe.lisp at gmail.com>:
>> Great, then why not replace ASDF's function user-homedir into this one:
>> 
>> (defun* user-homedir ()
>>  #+mcl
>>  (ccl::findfolder #$kuserdomain #$kCurrentUserFolderType)
>>  #-mcl
>>  (truenamize (pathname-directory-pathname (user-homedir-pathname))))
>> 
>> I've made a patch (in attach), include above new definition and some notes. I hope Faré could merge it.
>> 
> I'd do it, except the page you cite in your message suggests setting
> the binding for ccl::*user-homedir-pathname* instead. Shouldn't we do
> that instead? Or does any code rely on that file being the MCL path?
> Is the MCL path saved anywhere else? Note that user-homedir usually
> returns a truename, which translations depend upon to work properly.
> If you're hacking user-homedir instead of the global variable,
> shouldn't the conditional be inside the truename?
> http://code.google.com/p/mcl/wiki/Portability
> 
> Since you seem to be the one main user of MCL and ASDF, why don't you
> decide what it should be in the end? I'll release a 2.015.4 for you
> after you've made your mind.
> 
> —♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
> Speed has always been important otherwise one wouldn't need the computer.
>       —Seymour Cray