On Sun, Dec 16, 2001 at 10:56:20PM +0100, Hrvoje Niksic wrote: >Alan Eldridge <[EMAIL PROTECTED]> writes: > >> [...] >> DESCRIPTION >> These interfaces are obsoleted by random(3). >> >> I wrote the random-wait patch, and it used random(3) if that was >> available on the system. > >Yes, but if I remember correctly, your patch disabled the feature >completely if random was missing. rand() is always available, so I >switched to that. > >> We should not be using deprecated interfaces unless random() is not >> available. > >rand() may have its problems with reentrance and use in multithreaded >applications, but those are not relevant for Wget. rand() has made it >to ISO C, it's universally available, and it's certainly not going >anywhere.
If rand() is truly universally available, I have no problem. When my man page on FBSD tells me that the interface is obsolete, that raises a flag in my mind. I just checked SUSv2 and rand() and random() are both there. It would appear that random() obsoletes rand() because random() generates a better sequence. This is not something we are really concerned about here. "Good enough" is good enough. So, go with rand(), or, if multithreadedness (multiple simultaneous d/ls) ever may become a concern, rand_r(unsigned*). If rand() ever disappears on a platform, I'm sure you'll hear about it. :) -- Alan Eldridge Just another $THING hacker.