On Tue, November 29, 2016 20:47, Dmitry Boyarintsev wrote:
Hello Dmitry, > A question regarding your comment ( > http://bugs.freepascal.org/view.php?id=31023#c96393) > Specifically ad 2) > > Is it fine to safe-guard access to SpecialKey / ScanCode through the whole > crt unit? My only concern is that there're many places that has to be > adjusted making the patch rather big, (I'm looking into using > TRTLCriticalSection) I don't know what's the best solution, but unit Crt isn't thread-safe (not really supposed to be). If the handler runs in a different thread on MS Windows, the risk of breaking the event processing is IMHO rather high (well, probably partly depending on how badly the event loop is written - it's extremely high if there are no delays between event availability checks). Some alternative ideas coming to my mind: 1) Instead of using the special handler registered via SetConsoleCtrlHandler (invoked in a special thread), isn't it possible to process this event in the exception handler (running in the same thread probably)? 2) Instead of directly updating the ScanCode and SpecialCode variables (accessed directly during each and every ReadKey/KeyPressed call), maybe you could add a new variable specifically for this purpose which would signalize readiness of this special key event. ReadKey and KeyPressed could then check this variable if and only if they are not in the middle of combined key processing yet (before trying to read / poll additional events from the keyboard driver) and update SpecialKey and ScanCode themselves (inside the main thread already - then the update from the other thread becomes an atomic operation and thus probably safe). Maybe there are other options, I don't know. Tomas _______________________________________________ fpc-pascal maillist - fpc-pascal@lists.freepascal.org http://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-pascal
