b> > There are certainly no sparse files creatable with fseek and friends
> > on IRIX 4.
...
> Irix is based on SVR4 -- it was not written from scratch by SGI.
Minor nit -- he said Irix 4, which is SVR3 if I'm not mistaken.
> Every SVR4 box, indeed, every unix box I have used has had holes in files.
> I will admit to not having explicitly checked this particular feature on
> an SGI, but unless SGI has been removing documented SVID functionality
> for some reason, you are wrong.
>
> I will, of course, happily conceed to a reposting of a section of man
> page that explicitly states that the functionality has been removed.
Every Unix system on which I've thought to check has had sparse files
too. Not wanting to RTFM I tried the following program:
$ cat holes.c
#include <stdio.h>
main()
{
fseek(stdout, 1024*1024, SEEK_SET);
putchar('$');
}
On a BSDI system, here's what happened:
$ uname -a
BSD/386 ohare.Chicago.COM 1.1 0 i386
$ rm -f foo bar
$ ./holes > foo
$ cat foo > bar
$ ls -ls foo bar
1040 -rw-r--r-- 1 kls users 1048577 May 1 17:18 bar
16 -rw-r--r-- 1 kls users 1048577 May 1 17:18 foo
As seen in the -s field (the first one), foo really is sparse. I then
tried an Irix 4 system:
$ uname -a
IRIX Z0 4.0.1 11150233 IP4
$ rm -f foo bar
$ ./holes > foo
$ cat foo > bar
$ ls -ls foo bar
2049 -rw-r--r-- 1 kls ugserve 1048577 May 2 18:27 bar
2049 -rw-r--r-- 1 kls ugserve 1048577 May 2 18:27 foo
Much to my susprise, there do not appear to be any holes in foo. I
verified this with df, which indeed showed a change in space usage
consistent with a non-sparse file.
I'd try it on Irix 5.2, but we only have one system running that and
it appears to be down at the moment. I did try it on AIX 3.2.5, which
is more likely to be different from Unix than any other "Unix," and it
too appeared to have sparse files.
--
Karl Swartz |INet [EMAIL PROTECTED]
SLAC Computing Services | or [EMAIL PROTECTED]
1-415/926-3630 |UUCP uunet!lll-winken!unixhub!kls -or- ditka!kls
(SLAC and the US Dept. of Energy don't necessarily agree with my opinions.)