https://bz.apache.org/bugzilla/show_bug.cgi?id=57771
Michael Thorpe <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW --- Comment #6 from Michael Thorpe <[email protected]> --- (In reply to Eric Covener from comment #5) > While my "cgid handler" comment is invalid because this is the INCLUDES > filter calling into mod_cgid, I still don't see why there'd be any thread > change here. As a simple scenario, assume there are two MPM event threads: 1) Request A comes in and thread #1 starts handling it. 2) Request B comes in and thread #2 starts handling it. 3) Request A creates a CGI using conn_id=1. 4) Thread #1 puts request A to sleep while it waits for the CGI. 5) Request C comes in and thread #1 starts handling it (using conn_id=1). 6) Thread #2 finishes handling request B. 7) The CGI finishes, and thread #2 picks it up (changing its conn_id to 2) since thread #1 is still handling request C. 8) Request A finishes and thread #2 enters the cleanup_script function. 9) Thread #2 stuffs the *current* connection ID (in this case, 2) and asks the daemon for the PID for that conn_id, but that fails because no CGI was created with that conn_id. 10) Thread #2 prints an error message saying that it couldn't clean up the CGI for conn_id #1, because the error message code is correct even though the daemon-request code is wrong. > Can you capture "trace8" of this happening for a single request on a provided > simple SSI? On a single request you'll never see the problem. It's only when the server is loaded that requests can get moved around between threads. -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
