Re: [xml] newline in xmlTextWriter on win32

[Igor Zlatkovic <igor zlatkovic com>]

On 12/08/04 21:35, Michael Culbertson wrote:
> Hello,
>
> On Wed, 11 Aug 2004, Igor Zlatkovic wrote:
>
> > This conversion done by the C runtime has been deactivated in libxml on
> > purpose. The problem was that under some bizarre conditions,
> > xmlParseFile() followed by a xmlSaveFile() modified the data. It
> > replaced every \n with \r\n, even if it allready was a \r\n, ending up
> > with \r\r\n. One did the read-write again and it became \r\r\r\n.
>
>
> Hmm.. this is interesting since according to XML specification (as was
> pointed out recently in an other message on this list), xmlParseFile()
> should have converted all \r\n sequences in the input file to \n.  Then,
> when the file was output through xmlSaveFile() the C library would convert
> all of the single \n characters to the \r\n sequence.
>
> May I ask what sort of bizarre conditions caused the duplication?  The
> change to binary write mode occurred in version 1.107 of xmlIO.c (20 Jul
> 2003).  I checked the mailing list archive around that date and didn't see
> any discussion on this.  If there was any, could you point me in that
> direction?

That was long ago, strike me if I know the details still. Someone 
reported a mysterious multiplication of line ends and I could reproduce 
that with libxml which was current somewhere in late 2001 or early 2002.

But that wasn't the reason for the change to xmlio.c in 2003. The reason 
 for that was discussed on libxslt list, if I remember well. Could look 
into the archives, but there is no need. Look at this:

  <?xml version="1.0"?>
  <xsl:stylesheet
      xmlns:xsl="http://www.w3.org/1999/XSL/Transform";
      version="1.0">
    <xsl:output method="text"/>
    <xsl:template match="/">
      <xsl:text>&#xd;&#xa;</xsl:text>
    </xsl:template>
  </xsl:stylesheet>

Take this and transform any XML document with it, using xsltproc, and 
add the -o option to save the result in a file. The resulting file must 
be exactly two bytes long and contain the sequence \r\n, on all 
platforms. If libxml would allow the runtime to convert the newlines, 
the result would be a \r\r\n file on Windows, clearly making the 
processor a broken one, on that platform.

Ciao,
Igor