Remy Maucherat wrote: >> If we let the servlet wrapper deal with it - then we need a 'partial >> flush' >> in release() - which will just move the data to the parent buffer ( >> OutputBuffer from what I see ), but without calling flush on it. >> >> Remy, what do you think ? > > Yes, I think I agree. I rarely see any reason to flush, it just makes > things slower. I think it's a very good idea to see *all* the code from > Jasper 1 with a very critical eye and not assume it's there for a Good > Reason ;-) > > Did you test it ? Does it work good ? (I don't see why it wouldn't, just > wondering)
It seems ok - I'm more worried about the spec implications, i.e. is the flush() in relase() implied by some of the spec - or javadoc ? My reading is not, but others understand the spec better. For now I'll just change PageContextImpl.release to do flushBuffer in both cases ( i.e. included or not ). This will empty the page buffers in the OutBuffer. My reading of the javadoc is that release() shouldn't do that ( but discard it )- but that's what has been done in the past and is the least risky. If we agree on the release() behavior, then we need to move the flushBuffer() in the calling code ( the generated servlet ), before release() is called. -- Costin -- To unsubscribe, e-mail: <mailto:tomcat-dev-unsubscribe@;jakarta.apache.org> For additional commands, e-mail: <mailto:tomcat-dev-help@;jakarta.apache.org>
