https://bz.apache.org/bugzilla/show_bug.cgi?id=61164
--- Comment #6 from Mark Thomas <ma...@apache.org> --- The updated patch looks good. Thanks. I noticed that the import order has changed. If you could undo that it would be good but it isn't a big deal to fix when the patch is applied. Regarding tracking container vs client connection abort, I was looking at AbstractProcessor#setErrorState() around line 90. Having thought about this some more, a request attribute is a bit of a hack and I think I can see a batter way. What I was thinking was to also log 'X' if AbstractProcessor#getErrorState()#isIoAllowed() returns false. The correct way to access the Processor via the Request or Response is by defining an ActionCode and calling Request#acton() or Response#action(). Take a look at ActionCode#IS_ERROR and how it is used. I am thinking ActionCode#IS_IO_ALLOWED and then call Request#action() from the access log. I'm not sure if there is value in differentiating client and container aborts. I'm leaning towards not differentiating for now. If we add support for '%X' now that does not differentiate, we can always add '%x' at a later date that does differentiate if there is demand. -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
