|
From: Julian S. <js...@ac...> - 2005-10-17 22:38:16
|
As a general point, the fact that Purify reports no error and V does does not necessary mean that V is wrong. In reality both systems have to approximate reality to some extent. It's not difficult to write a program with an error which Purify misses but Valgrind doesn't. It's arguable that V does more accurate value tracking than Purify and experience tends to show that, for the most part, when it complains about something it is correct. Sure, it's possible to fool it into reporting nonexistent problems re uses of uninitialised variables, but for real production code (as opposed to carefully constructed test cases) but this tends to be fairly rare. J |