• last updated a few seconds ago
Constraints: committers
Constraints: files
Constraints: dates
Port WS patch: upgrade behavior tweaks.
Name cleanup (flush became a blocking write of the remaining data).
Port patch for TC BZ56010: avoid IAE when using the constant for the default value.
Attempt another autoblocking technique. Uses a fake blocking write notification that would write all content and flush.
Obvious cleanup: don't wrap as an IOE if already a meaningful IOE.
BZ1027272 (part 1): Attempt to set the CCL and thread binding when expiring sessions from other contexts.
Port the alwaysUseSession option from Tomcat, disabled by default (as in Tomcat).
    • -0
    • +16
Use my patch from Tomcat instead: don't do anything at all if nothing was written yet.
JBWEB-287: Add a null check because the write event should be called even if no write was made (from the Servlet spec).
Web 2.1.14.
Web 2.1.14.
Drop the broken lazy init in favor of a simple flag to indicate if websockets is used.
Port websockets patches.
BZ 1030050: Sync FORM fixes with 7.4 for saving request body when using FORM.
Port patches: Fix possible NPE on destroy. Update filter name.
Reset the (kinda new) EOF flag when restoring the request body.
Some (basic) constant cleanups and removal of old unused code.
Add the configuration system properties used in the other HTTP connectors.
Port improvements to content-length handling.
Arrange doc and make sure the JSSE behaviour mimics the openssl one.

    • -1
    • +1
Port minor addition to allow replacing SSL context.
BZ 1030050: Sync some changes that offer more flexibility (in addition to using createInputStream which allows using a reader for the restored request).
BZ1036977: Fix using modeler (bad mbean names) with native and NIO2 connectors.
Fix apparently badly ported JMX code (java.io is correct, but not NIO 2 or native). Missed AJP.
Compatibility with new JDT.
Fix apparently badly ported JMX code (java.io is correct, but not NIO 2 or native).
Port BZ 1035787: Submitted by Ladislav Thon, once paused (and not resumed) the connector would leak some threads. Visible during testing management mostly.