Hi! > I don't wish to over complicate this, but how many potential fd types > could there be 100, 1000? Some could have complicated init logic. I'm at 25 at the moment, suprisingly all of them so far are a syscall with a few parameters, sometimes packed in a struct. > I'm wondering if at the outset it would be better to define an interface > struct with name, setup and teardown for each FD type, plus whatever > other meta-data might be useful for filtering. > > Then instead of a case statement, we put the structs in an array etc. I guess that we can, but we would have to add some private data area to the tst_fd, so that we can tear down things cleanly, but we would need that if we want to convert the tst_iterate_fd() to be iterator-like anyways. -- Cyril Hrubis chrubis@xxxxxxx
