Edit report at https://bugs.php.net/bug.php?id=63379&edit=1
ID: 63379 Comment by: avatar2004-php at yahoo dot fr Reported by: avatar2004-php at yahoo dot fr Summary: Warning when using session_regenerate_id(TRUE) with a SessionHandler Status: Assigned Type: Bug Package: Session related Operating System: Gentoo PHP Version: 5.4.8 Assigned To: arpad Block user comment: N Private report: N New Comment: Looking at the code, it feels like the very semantics of the mod_user_is_open flag are not exactly consistent. The flag is a global, yet it is manipulated by instance code. Meaning if the user space code uses several handlers interchangeably, the result can quickly become confusing. I suggest a decision should be made as to whether the SessionHandler is a stateless, thin wrapper for the handler calls in which case it should default to the same checks as the procedural API. Otherwise, the flag should really be an instance variable used to track the proper invocation sequence of the different callbacks and make sure THIS handler is open before calling dependent routines. Adding an instance variable to the (base) class though is probably not worth it. The session management semantics are already defined by the procedural API which uses a global session state flag (session_status). If required for a handler implementation, the flag can simply be implemented in user space. Just my 2c. Previous Comments: ------------------------------------------------------------------------ [2012-10-29 03:07:41] larue...@php.net I mean, maybe only reset it in close handler? ------------------------------------------------------------------------ [2012-10-29 03:07:08] larue...@php.net is the reseting of user_is_open necessary? diff --git a/ext/session/mod_user_class.c b/ext/session/mod_user_class.c index 70d2f40..4edac28 100644 --- a/ext/session/mod_user_class.c +++ b/ext/session/mod_user_class.c @@ -121,7 +121,6 @@ PHP_METHOD(SessionHandler, destroy) return; } - PS(mod_user_is_open) = 0; RETVAL_BOOL(SUCCESS == PS(default_mod)->s_destroy(&PS(mod_data), key TSRMLS_CC)); } /* }}} */ ------------------------------------------------------------------------ [2012-10-29 01:14:11] avatar2004-php at yahoo dot fr If I understand correctly and "mod_user_is_open" is just a global state parameter used by the SessionHandler instance, I was wondering why the check wasn't being done on "session_status" instead to conform with the rest of the session_* API ? ------------------------------------------------------------------------ [2012-10-29 01:04:34] avatar2004-php at yahoo dot fr Description: ------------ It seems there's an issue with the SessionHandler implementation and the way the destroy handler is invoked when using session_regenerate_id(TRUE) The problem seems to come from using the global "mod_user_is_open" in the PS_SANITY_CHECK_IS_OPEN macro (in ext/session/mod_user_class.c). The test script generates an undue warning: Test script: --------------- $handler = new SessionHandler(); session_set_save_handler($handler, TRUE); // or FALSE, doesn't matter session_start(); session_regenerate_id(TRUE); //session_write_close(); Expected result: ---------------- No warning Actual result: -------------- PHP Warning: Unknown: Parent session handler is not open in Unknown on line 0 ------------------------------------------------------------------------ -- Edit this bug report at https://bugs.php.net/bug.php?id=63379&edit=1