From xach at xach.com  Mon Jun 13 19:51:22 2011
From: xach at xach.com (Zach Beane)
Date: Mon, 13 Jun 2011 15:51:22 -0400
Subject: [closure-devel] Use of (make-pathname ... :type :unspecific)
Message-ID: <87boy17bnp.fsf@hangup.portland.xach.com>

In src/parse/html-parser.lisp, there's this form:

    (make-pathname :directory '(:relative "resources") :type :unspecific)

19.2.2.2.3 says this:

    Whether a value of :unspecific is permitted for any component on any
    given file system accessible to the implementation is
    implementation-defined. A conforming program must never
    unconditionally use a :unspecific as the value of a pathname
    component because such a value is not guaranteed to be permissible
    in all implementations.

CLISP signals an error on the form.

What's the :type :unspecific for? It seems to work on CLISP (and all
other implementations I tried) without it; is it a precaution against a
*default-pathname-defaults* that apparently has a value for :type?

Zach



From david at lichteblau.com  Mon Jun 13 20:18:18 2011
From: david at lichteblau.com (David Lichteblau)
Date: Mon, 13 Jun 2011 22:18:18 +0200
Subject: [closure-devel] Use of (make-pathname ... :type :unspecific)
In-Reply-To: <87boy17bnp.fsf@hangup.portland.xach.com>
References: <87boy17bnp.fsf@hangup.portland.xach.com>
Message-ID: <20110613201818.GE7188@radon>

Quoting Zach Beane (xach at xach.com):
[...]
> What's the :type :unspecific for? It seems to work on CLISP (and all
> other implementations I tried) without it; is it a precaution against a
> *default-pathname-defaults* that apparently has a value for :type?

I don't know.  Here's the snippet from the patch submitter, can you make
sense of it?  Perhaps only very old asdf2 has this problem?

* Slava Gorbunov wrote on 20 Sep 2010:
| And, finally, here is patch which makes possible to load closure-html
| with asdf2 (included in recent versions of SBCL, for example):
|
| --- a/src/parse/html-parser.lisp
| +++ b/src/parse/html-parser.lisp
| @@ -34,7 +34,7 @@
|  (defparameter sgml::*simple-catalog*
|    (let ((base
|          (merge-pathnames
| -         "resources/"
| +         (make-pathname :directory '(:relative "resources") :type :unspecific)
|           (asdf:component-relative-pathname
|            (asdf:find-system :closure-html)))))
|      (loop
|
| asdf:component-relative-pathname in ASDF2 returns pathname with .asd
| extension appended, so open-public-resource tries to open files with
| .asd extension appended, too, and, obviously, fails.


d.



From xach at xach.com  Mon Jun 13 21:59:30 2011
From: xach at xach.com (Zach Beane)
Date: Mon, 13 Jun 2011 17:59:30 -0400
Subject: [closure-devel] Use of (make-pathname ... :type :unspecific)
In-Reply-To: <20110613201818.GE7188@radon> (David Lichteblau's message of
	"Mon, 13 Jun 2011 22:18:18 +0200")
References: <87boy17bnp.fsf@hangup.portland.xach.com>
	<20110613201818.GE7188@radon>
Message-ID: <877h8p75q5.fsf@hangup.portland.xach.com>

David Lichteblau <david at lichteblau.com> writes:

> Quoting Zach Beane (xach at xach.com):
> [...]
>> What's the :type :unspecific for? It seems to work on CLISP (and all
>> other implementations I tried) without it; is it a precaution against a
>> *default-pathname-defaults* that apparently has a value for :type?
>
> I don't know.  Here's the snippet from the patch submitter, can you make
> sense of it?  Perhaps only very old asdf2 has this problem?

Yes, that looks like it's the case to me. ASDF2 no longer behaves like
he describes.

Zach



> * Slava Gorbunov wrote on 20 Sep 2010:
> | And, finally, here is patch which makes possible to load closure-html
> | with asdf2 (included in recent versions of SBCL, for example):
> |
> | --- a/src/parse/html-parser.lisp
> | +++ b/src/parse/html-parser.lisp
> | @@ -34,7 +34,7 @@
> |  (defparameter sgml::*simple-catalog*
> |    (let ((base
> |          (merge-pathnames
> | -         "resources/"
> | +         (make-pathname :directory '(:relative "resources") :type :unspecific)
> |           (asdf:component-relative-pathname
> |            (asdf:find-system :closure-html)))))
> |      (loop
> |
> | asdf:component-relative-pathname in ASDF2 returns pathname with .asd
> | extension appended, so open-public-resource tries to open files with
> | .asd extension appended, too, and, obviously, fails.
>
>
> d.



From david at lichteblau.com  Tue Jun 14 14:26:13 2011
From: david at lichteblau.com (David Lichteblau)
Date: Tue, 14 Jun 2011 16:26:13 +0200
Subject: [closure-devel] Use of (make-pathname ... :type :unspecific)
In-Reply-To: <877h8p75q5.fsf@hangup.portland.xach.com>
References: <87boy17bnp.fsf@hangup.portland.xach.com>
	<20110613201818.GE7188@radon>
	<877h8p75q5.fsf@hangup.portland.xach.com>
Message-ID: <20110614142613.GF7188@radon>

Quoting Zach Beane (xach at xach.com):
> David Lichteblau <david at lichteblau.com> writes:
> 
> > Quoting Zach Beane (xach at xach.com):
> > [...]
> >> What's the :type :unspecific for? It seems to work on CLISP (and all
> >> other implementations I tried) without it; is it a precaution against a
> >> *default-pathname-defaults* that apparently has a value for :type?
> >
> > I don't know.  Here's the snippet from the patch submitter, can you make
> > sense of it?  Perhaps only very old asdf2 has this problem?
> 
> Yes, that looks like it's the case to me. ASDF2 no longer behaves like
> he describes.

Does this change work for you on clisp?

--- a/src/parse/html-parser.lisp
+++ b/src/parse/html-parser.lisp
@@ -33,10 +33,13 @@
 
 (defparameter sgml::*simple-catalog*
   (let ((base
-        (merge-pathnames
-         (make-pathname :directory '(:relative "resources") :type :unspecific)
-         (asdf:component-relative-pathname
-          (asdf:find-system :closure-html)))))
+        (make-pathname
+	 :name nil
+	 :type nil
+	 :defaults (merge-pathnames
+		    "resources/"
+		    (asdf:component-relative-pathname
+		     (asdf:find-system :closure-html))))))
     (loop
        :for (name . filename)
        :in '(("-//W3O//DTD W3 HTML 3.0//EN" . "dtd/HTML-3.0")



From xach at xach.com  Tue Jun 14 14:29:04 2011
From: xach at xach.com (Zach Beane)
Date: Tue, 14 Jun 2011 10:29:04 -0400
Subject: [closure-devel] Use of (make-pathname ... :type :unspecific)
In-Reply-To: <20110614142613.GF7188@radon> (David Lichteblau's message of
	"Tue, 14 Jun 2011 16:26:13 +0200")
References: <87boy17bnp.fsf@hangup.portland.xach.com>
	<20110613201818.GE7188@radon>
	<877h8p75q5.fsf@hangup.portland.xach.com>
	<20110614142613.GF7188@radon>
Message-ID: <87pqmg4hcf.fsf@hangup.portland.xach.com>

David Lichteblau <david at lichteblau.com> writes:

> Quoting Zach Beane (xach at xach.com):
>> David Lichteblau <david at lichteblau.com> writes:
>> 
>> > Quoting Zach Beane (xach at xach.com):
>> > [...]
>> >> What's the :type :unspecific for? It seems to work on CLISP (and all
>> >> other implementations I tried) without it; is it a precaution against a
>> >> *default-pathname-defaults* that apparently has a value for :type?
>> >
>> > I don't know.  Here's the snippet from the patch submitter, can you make
>> > sense of it?  Perhaps only very old asdf2 has this problem?
>> 
>> Yes, that looks like it's the case to me. ASDF2 no longer behaves like
>> he describes.
>
> Does this change work for you on clisp?
>
[snip]

CLISP compiles it without complaint with that change.

Zach



From xach at xach.com  Tue Jun 14 15:28:03 2011
From: xach at xach.com (Zach Beane)
Date: Tue, 14 Jun 2011 11:28:03 -0400
Subject: [closure-devel] Use of (make-pathname ... :type :unspecific)
In-Reply-To: <20110614142613.GF7188@radon> (David Lichteblau's message of
	"Tue, 14 Jun 2011 16:26:13 +0200")
References: <87boy17bnp.fsf@hangup.portland.xach.com>
	<20110613201818.GE7188@radon>
	<877h8p75q5.fsf@hangup.portland.xach.com>
	<20110614142613.GF7188@radon>
Message-ID: <87hb7s4em4.fsf@hangup.portland.xach.com>

Thanks for committing the update. An unrelated update introduces an
implicit dependency on CXML by way of CXML::VALID-NAME-P; I get "No
package named CXML" as an error now unless CXML is coincidentally
already loaded.

Zach