2014-10-14 16:17 GMT+04:00 Konstantin Kolinko <knst.koli...@gmail.com>: > 2014-10-14 10:12 GMT+04:00 Keiichi Fujino <kfuj...@apache.org>: >> I have a question of new SessionIdGenerator implementation. >> If reloading Manager( or Context), SessionIdGenerator is re-created based >> on sessionIdGeneratorClass that is cached to Manager. >> However, attributes of jvmRoute and sessionIdLength are default value. >> I think these attributes are need to take over when reloading. >> Or another workaround? > > Looking at ManagerBase.startInternal(), I think that > > 1) jvmRoute shall not be set on <SessionIdGenerator> element. It is > set on <Engine> and is inherited from there. The documentation on > "jvmRoute" attribute in config/sessionidgenerator.html explicitly > mentions that effect. > > It ManagerBase.startInternal() there is the following code: > "sessionIdGenerator.setJvmRoute(getJvmRoute());" > > 2) For sessionIdLength I think one shall be able to configure it on > <SessionIdGenerator>, so the current behaviour is a bug and is worth > to be added to Bugzilla. > > The workaround is to set that attribute on <Manager> element. (This > contradicts documentation that says that this attribute on Manager > element is deprecated). > > The fix is probably to let SessionIdGenerator to implement Lifecycle > interface. > > The "this.sessionIdGenerator = null;" assignment in > ManagerBase.stopInternal() originates from r1044874. That is a wrong > place to clear sessionIdGenerator reference. A better place would be > destroyInternal(), or just do not clear it. > > The old code before r1044874 was calling "randoms.clear();" there. I > think it is easier to expose that method via Lifecycle.stop().
I filed https://issues.apache.org/bugzilla/show_bug.cgi?id=57089 Best regards, Konstantin Kolinko --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
