Patch for ASDF 3

The way you look for the path of catalog.dtd doesn't work in ASDF 3.
Here is a supported way that works with all versions of ASDF 2 and later.

I'll try to improve backward compatibility, but you may nonetheless
want to apply the patch.
Also, I believe that three years afterwards, and
with each and every implementation providing ASDF 2,
it is safe to remove the #-asdf2 code around your codebase.

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
Work harder! Millions on welfare depend on you.

Possible bug in rune-dom::adjust-vector-exponentially

Hi,

I have a process that is stuck in the following function:

;; from dom-impl.lisp
(defun adjust-vector-exponentially (vector new-dimension set-fill-pointer-p)
  (let ((d (array-dimension vector 0)))
    (when (< d new-dimension)
      (loop
	  do (setf d (* 2 d))
	  while (< d new-dimension))
      (adjust-array vector d))
    (when set-fill-pointer-p
      (setf (fill-pointer vector) new-dimension))))

Apparently, adjust-vector-exponentially assumes d > 0, but this function has been called from (METHOD
SAX:CHARACTERS (RUNE-DOM::DOM-BUILDER T)) with an empty string (d = 0).  The loop obviously shows it will
never exit the function.

Adding the following before the loop statement fixes the problem:

      (when (= d 0)
        (incf d))

Best Regards,
Camille

Incorrect character with code 12 in output?

Hello,

I have a database with text data which sometimes is not proper to valid XML.
For example, there is a field sometimes containing the character new-page or ^L in emacs (ASCII code = 12).

While CXML does not complain and outputs cmxl:text as it was given, another XML parser complains that this
is not valid XML.

	PCDATA invalid Char value 12

I am expecting CXML to produce valid output, is there something I missed to make it more strict?

I am using LispWorks 64 bits.

Thanks!

Best Regards,
Cam

Klacks bug with &# entities on CCL

Hi there,

The following code breaks in CCL (Version 1.8-r15286M  (DarwinX8632))
with the most recent cxml from quicklisp (cxml-20110619-git):

(klacks:with-open-source (source (cxml:make-source "<value>&#32;987lkjasd</value>"))
  (klacks:consume source)
  (klacks:consume source)
  (klacks:consume source))

Here's the SLIME backtrace I get from evaluating that form:

  0: ((SETF RUNES:%RUNE) #\  "" 0)
  1: (CXML::READ-TOKEN-3 #S(CXML::ZSTREAM :TOKEN-CATEGORY NIL :TOKEN-SEMANTIC NIL :INPUT-STACK ...))
  2: (CXML::PEEK-TOKEN #S(CXML::ZSTREAM :TOKEN-CATEGORY NIL :TOKEN-SEMANTIC NIL :INPUT-STACK ...))
  3: (CXML::KLACKS/CONTENT #<CXML::CXML-SOURCE #xD7DE306> #S(CXML::ZSTREAM :TOKEN-CATEGORY NIL
:TOKEN-SEMANTIC NIL :INPUT-STACK ...) #<CCL:COMPILED-LEXICAL-CLOSURE (:INTERNAL ..))
  4: (CXML::KLACKS/ELEMENT-2 #<CXML::CXML-SOURCE #xD7DE306> #S(CXML::ZSTREAM :TOKEN-CATEGORY NIL
:TOKEN-SEMANTIC NIL :INPUT-STACK ...) (("") ("xmlns" . "http://www.w3.org/2000/xmlns/") ..) #<CCL:COMPILED-L..
  5: (CXML::FILL-SOURCE #<CXML::CXML-SOURCE #xD7DE306>)
  6: (#<STANDARD-METHOD KLACKS:CONSUME (CXML::CXML-SOURCE)> #<CXML::CXML-SOURCE #xD7DE306>)
  7: (#<Anonymous Function #xD7E474E>)
  8: (CCL::CHEAP-EVAL (KLACKS:WITH-OPEN-SOURCE (SOURCE (MAKE-SOURCE
"<value>&#32;987lkjasd</value>")) (KLACKS:CONSUME SOURCE) (KLACKS:CONSUME SOURCE)
(KLACKS:CONSUME SOURCE)))

That same form works on SBCL, so I assume there's some
implementation-specific charset-related weirdness going on. I'm happy to
help debug this, but sadly, my code-diving time is a bit limited at the
moment.

Bug in DTD parsing?

(flex:with-input-from-sequence (s (babel:string-to-octets
                                   "<![IGNORE[ ]]><!-- -->"
                                   :encoding :utf-8))
  (cxml:parse-dtd-stream s))

errors with
Document not well-formed: Unexpected character #\!.
Context:
  Line 1, column 17 in NIL

   [Condition of type CXML:WELL-FORMEDNESS-VIOLATION]

but parses OK with whitespace between ]]> and <!--

(for a real example, see line 396 in http://docbook.org/xml/4.3/dbpoolx.mod)

it looks like P/IGNORE-SECT in cxml/xml/xml-parse.lisp should unread 1
rune after matching ]]> (or exit the loop sooner so it doesn't read an
extra).

Inconsistency in creating %children vector

I was scratching my head for a couple of hours today, because of an apparent failure of
cxml-dom::make-space - it did not increase the size of the %children vector. On closer examination, it
turned out that the reason was that the %children vector was not adjustable, at that point.

The only places in cxml-stp where %children is set up are in %unchecked-insert-child (parent-node.lisp)
and normalize-text-nodes! in xpath.lisp. The first of these has (make-array ... :adjustable t ...),
while the other does not. My guess is that normalize-text-nodes! might be called as part of an xpath
evaluation, and if this happens in code that is actually building the cxml-stp structure, the result
might be a situation where the %children vector is not adjustable.

I'm not sure if this is a valid use of cxml-stp/xpath, but there is at least one documented case where it has
been done 

I could (trivially) change my code to not use xpath as part of the construction process, but adding
:adjustable t to the call to make-array in xpath.lisp makes my (current) code work, and I guess it also
makes cxml-stp slightly more robust.

Real-world example: with attachment

Real-world Example

namespace-uri usage

cxml-stp usage

Exported symbols [was: Re: Possible bug in cxml-stp]