Re: Proposal for the ISO character entity bug in gnujaxp

Arnaud Vandyck <[email protected]> writes:

> Hi everybody, hi Nic,
> 
> Looking at you problem with your xml fragment
> <http://mail.gnu.org/pipermail/classpathx-xml/2002-May/000125.html>, I
> did some tests to find where is the bug.
> 
> First, I created a test class to show the nodes of a document with
> DOM. I also did an xml file near yours and I saw you were working with
> DocBook for xml (SimpleDocbook if I recall). I did some tests and
> found that the problem was when parsing the ISO character entity of
> Docbook (for example: <!ENTITY euro "&#x20AC;">).
> 
> I did change the call of the readLiteral(int) method in the
> parseEntityDecl() method in the gnu.xml.aelfred2.XmlParser class.
> 
> This call is to get the value of an entity if I did understant the
> code correctly. But it passes the value '0'. I did change it to value
> LIT_DISABLE_CREF (this is an integer, the value is 16). Looking in
> other methods, I saw the developers use an 'int flags=LIT_DISABLE_...'
> line. I did the same in the parseEntityDecl() method.
> 
> I do not know if this will not cause any other problems and I suggest
> everyone who can to look at the patch before committing the
> changes... maybe someone can forward this mail to one of the author or
> both (David Megginson <[email protected]>, David Brownell
> <[email protected]>) to be sure the patch is correct.
> 
> Please boys, be indulgent for my first patch ever! :-) If I gave to
> many descriptions (too verboze?), if the patch format is not good, if
> there is a problem, let me know. If my english is not very good, well,
> don't let me know, I already do :-)
> 
> These commands patch the good file:
> 
> # cd jaxp_root
> # patch -p1 < ../jaxp-char-entities-patch-20021012.diff
> 
> This is also the first time I use the diff and patch commands!
> 
> I hope this helps, have a nice day/night :-)
> 
> PS: I also attached a Test.java and a test.xml files.
>     You'll have to change the dtd location, I used the docbook xml
>     v4.2 for the test.

Just to let you know... I'm doing this now. It looks ok so I'll
probably check it in tonight.

I'm going to check in the test code into a new test framework too.


Nic