Hi all, So I _think_ I've fixed this bug. In the AbstractTextSerializer, the setOutputStream method wraps the incoming output stream in a BufferedOutputStream. However, no class reference is kept to this BOS, so it never gets flushed at the end of the document. I made the BOS a class variable (instead of a local method variable), and added an endDocument() implementation that flushes the BOS and sets it to null (and then invokes the superclass's endDocument() method). This seems to fix the problem.
What I still don't understand, though, is why this only occurred when I used the ASCII encoding. I'm guessing this is because I don't fully understand how the serializers handle encoding, but if anyone can provide some insights, that would be great. Thanks! Harry -----Original Message----- From: Lai, Harry [mailto:[EMAIL PROTECTED]] Sent: Tuesday, July 02, 2002 2:30 PM To: '[EMAIL PROTECTED]' Subject: RE: [Possible Bug] TextSerializer loses characters with ASCII enc odin g? I should also mention that I'm using Saxon 6.5 as my XSLT engine. Harry -----Original Message----- From: Lai, Harry [mailto:[EMAIL PROTECTED]] Sent: Tuesday, July 02, 2002 2:19 PM To: 'Cocoon Users' Subject: [Possible Bug] TextSerializer loses characters with ASCII encodin g? Hi folks, I'm seeing some really odd behavior with serializers, and wanted to see if anyone had seen this before (or better yet) knows what's causing this. =) PROBLEM: If I set an XML, HTML, or Text serializer's encoding to ASCII, I end up losing the tail end of my output (roughly 1-1000 characters). I think it has to do with the buffered output stream not being properly closed/flushed, but I'm not sure. ENVIRONMENT: Cocoon 2.0.2 running on JBoss 2.4.5 with Tomcat 4.0.3. Win2k. WORKAROUNDS: I can make this problem go away by doing either of the following: 1. Change the encoding to UTF-8 or ISO-8859-1. 2. Change the buffer-size to 1 (tried with values from 1-2048; was still seeing the problem with values as low as 10). REPRODUCING THE BUG: >From what I'm seeing, you can reproduce this bug by having a reasonably large buffer (larger than 10 bytes) and switching any TextSerializer-based serializer's encoding to ASCII. Here's a sample from my sitemap: <map:serializer name="css" mime-type="text/css" logger="sitemap.serializer.css" src="org.apache.cocoon.serialization.TextSerializer" pool-max="64" pool-min="16" pool-grow="4"> <buffer-size>2048</buffer-size> <encoding>ASCII</encoding> </map:serializer> Any thoughts, suggestions, ideas, etc. would be greatly appreciated. Thanks! Harry PS The main reason this comes up is for a dynamic css pipeline, which I'd like to be character set agnostic (hence, the ASCII encoding). I only tested it with the HTML and XML serializer to make sure it wasn't something specific to my css pipeline. -------------- Harry Lai [EMAIL PROTECTED] 512.682.1101 x2170 --------------------------------------------------------------------- Please check that your question has not already been answered in the FAQ before posting. <http://xml.apache.org/cocoon/faq/index.html> To unsubscribe, e-mail: <[EMAIL PROTECTED]> For additional commands, e-mail: <[EMAIL PROTECTED]> --------------------------------------------------------------------- Please check that your question has not already been answered in the FAQ before posting. <http://xml.apache.org/cocoon/faq/index.html> To unsubscribe, e-mail: <[EMAIL PROTECTED]> For additional commands, e-mail: <[EMAIL PROTECTED]> --------------------------------------------------------------------- Please check that your question has not already been answered in the FAQ before posting. <http://xml.apache.org/cocoon/faq/index.html> To unsubscribe, e-mail: <[EMAIL PROTECTED]> For additional commands, e-mail: <[EMAIL PROTECTED]>
