On Tue, 2010-09-14 at 07:50 +0200, Mladen Turk wrote: > On 09/14/2010 12:28 AM, Brett Delle Grazie wrote: > >> > >> Detaching AJP streams from the servlet and using them > >> as a OutputStream is not very well handled in Tomcat. > >> I suppose if using APR the entire thing would even crash the JVM. > >> > >> Trying to deal with that on mod_jk side would have my strong -1. > > > > And I would agree with that whole heartedly. > > > > So I guess the APR AJP connector implementation needs to be reviewed... > > I am looking but as I'm not familiar with the code any help or pointers > > would be appreciated. > > > > For AJP APR connector 'if (actionCode == ActionCode.ACTION_CLIENT_FLUSH)' > doesn't check if the socket was already virtually closed > so I guess that would be a first thing to look at. > We cannot just check for finished flag thought, cause recycle() would reset > that. > Instead if there was getOutputSteam() we would need to directly > invalidate that stream. This means we would need to remember that stream > somewhere. > > Further more the wrapped socket is 'long' pointing to the > native pointer, so we don't have object notification on > physical socket close (that's another problem that manifests > in JVM core for detached sockets). > > > Should I raise a bugzilla bug? > > > > Yes, and a dev list would be better place to discuss that further.
Hi, Bug raised: https://issues.apache.org/bugzilla/show_bug.cgi?id=49929 #49929 Thanks, > > > Regards -- Best Regards, Brett Delle Grazie ______________________________________________________________________ This email has been scanned by the MessageLabs Email Security System. For more information please visit http://www.messagelabs.com/email ______________________________________________________________________ --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
