Subject: Re: signal(SIGSEGV, SIG_IGN) -> 100% CPU
To: Gandhi woulda smacked you <greywolf@starwolf.com>
From: Robert Elz <kre@munnari.OZ.AU>
List: tech-kern
Date: 06/15/1999 07:21:02
    Date:        Mon, 14 Jun 1999 13:45:47 -0700 (PDT)
    From:        Gandhi woulda smacked you <greywolf@starwolf.com>
    Message-ID:  <Pine.NEB.4.02.9906141331550.2687-100000@starwolf.starwolf.com>

  | Talk about apples and oranges!  Try replacing, in the above
  | comparison, write, brk/sbrk or mmap.

Huh?   All I was saying is that Greg's "system calls can't be replaced,
but other libc() routines can" was bogus.   Some sys calls are important
for the library (including those you mention), some aren't.  The same is
true of other library routines.

  | I don't know about anyone else, but if I'm going to be writing my own
  | library of stuff, chances are, unless it's a graphics library, I'll be
  | using system calls all over the place because context switches, to my
  | mind, aren't cheap.

If I were doing that, I'd be profiling the thing, and then attempting to
fid ways to optimise the slow parts.   If there's a routine I'm calling
which is slow (whatever the reason) and it mattered to the results, then
I'd look and see if I can find a way to avoid it.   Whether that routine
happens to be a sys call or not would seem to be irrelevant.

  | Besides, man(1) figures out which section it's in; if you don't care
  | which section it's in, let man  do its job :-)

Have you tried "man mount" recently?

kre