Messages in this thread | | | Date | Thu, 22 Oct 1998 20:23:00 -0500 | From | Mitchell Blank Jr <> | Subject | Re: [patch] jiffies wraparound [Re: 2.1.125 Show stopper list: Draft] |
| |
Linus Torvalds wrote: > You can get a lot of that from /proc already.
Yes, but: 1. /proc requires parsing. Do we really win if we save 50 bytes from the kernel at the cost of 5000 bytes in libc? Yes, libc is swappable but they both need to fit on the disk. 2. /proc isn't universally available. Not neccesarily a big deal. Just hope that nothing early in your rc scripts need correct sysconf values. 3. Worse, if /proc needs to be mounted in every chroot you use if you want sysconf to work.
Building sysconf peicemeal out of /proc is A) broken and B) ugly.
> Quite frankly, I don't care about theory. Yes, in theory there's a lot you > need. In practice, I don't see any real people clamouring for it.
In theory we're tring to get closer to UNIX98 compliance.
> And discussing this at this point is useless anyway. It's not getting past > the code-freeze - if people haven't been needing it before now, there's no > excuse for saying that we need it NOW.
And I respect that. That is _fine_. The only reason I mention it now is because it's getting in my way now. However, it would be nice if you could at least think about it, realize that this is the correct approach, and put in sys_sysconf, sys_pathconf, and sys_fpathconf as ENOSYS stubs. That way things like glibc and m68k-native-patch could start targeting that interface. Or if you have some other idea on how it should be implemented, tell us so we can do it that way. Please rethink your do-it-in-libc-at-all-costs stance, though.
-Mitch
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |