On Tue, 2008-10-07 at 20:51 +0100, sebb wrote: ...
> > > > Modified: > > httpcomponents/httpcore/trunk/module-main/src/main/java/org/apache/http/impl/io/ChunkedInputStream.java > > URL: > > http://svn.apache.org/viewvc/httpcomponents/httpcore/trunk/module-main/src/main/java/org/apache/http/impl/io/ChunkedInputStream.java?rev=702599&r1=702598&r2=702599&view=diff > > > > ============================================================================== > > --- > > httpcomponents/httpcore/trunk/module-main/src/main/java/org/apache/http/impl/io/ChunkedInputStream.java > > (original) > > +++ > > httpcomponents/httpcore/trunk/module-main/src/main/java/org/apache/http/impl/io/ChunkedInputStream.java > > Tue Oct 7 12:33:57 2008 > > @@ -219,8 +219,7 @@ > > this.buffer.clear(); > > int i = this.in.readLine(this.buffer); > > if (i == -1) { > > - throw new MalformedChunkCodingException( > > - "Chunked stream ended unexpectedly"); > > Perhaps log a message here, as it is non-standard? > > > + return 0; > > } > > int separator = this.buffer.indexOf(';'); > > if (separator < 0) { > > HttpCore (for better or worse) does not have a dependency on a logging toolkit. One possibility would be introducing a strict / lenient flag and throwing MalformedChunkCodingException in the strict mode. However, this cannot be done without breaking API compatibility with Android, not until the 4.0.1 release. Besides, as it turned out the non-blocking chunk content decoder in HttpCore NIO has always been lenient about the missing closing chunk. At the very least, both implementations, blocking and non-blocking, are consistent now. I think it is a fairly harmless protocol violation, which probably can be tolerated, as it should not result in any data corruption. Oleg --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
