Aside from putting on kernel 2.0.39-pre4 and DAC960 support,
there are new versions of pax and dd.  The dd especially I'd
like tested, and I'd like some feedback from other users on:

What output do you get on _your_ system from:

 dd if=/dev/zero bs=10000 count=1 > test1
 dd if=/dev/zero bs=10000 count=1 | dd bs=10000 count=1 > test2
 ls -l test1 test2

Now, to MY lights, dd should be usable this way to count a chunk of a
file, it should NOT depend on the size of the shell's pipe buffer, but,
on all the 'dd' version I can find (except the one I patched for .204),
the second 'dd' counts 1 input block based on what is available in the
pipe, and the results are different.  Well, I have to admit, it seems
to be "broken" on HP/UX (at 8192) and Solaris (at 10240), also, but,
given that you never know exactly what a pipe will buffer, it makes it
useless for counting/skipping/seeking unless you know exactly what the
physical characteristics are of what you are reading from.

Now, it turns out that the only reason I included 'dd' on tomsrtbt in
the FIRST place was I wanted an accurate way of doing seek/skip/count
stuff, I certainly don't want to be forced to using "bs=1 count=10000",
yes, that would work, but is painfully inefficient, I don't really want
to read one byte at a time from the floppy.

_SO_, what I have done, is changed the "safe_read" routine in the
fileutils/lib for the "dd" so that it will keep trying to fill up
the requested read until it either has a full buffer, or EOF, or an
error.  This means some inefficiencies if you are trying to use it
just instead of the 'buffer' program, I suppose, but what I really
want to know is, can anyone think of any other negative side effects?

-Tom

Reply via email to