yes, works here too that way.. hmmm.
(quote from strace sh):
write(1, "1", 1) = 1
write(1, " 2", 2) = 2
write(1, " 3", 2) = 2
write(1, "\n", 1)
(quote from strace /bin/echo):
write(1, "1 2 3\n", 6)
->
This used to work, I suspect "bash" changed policy....
So I guess the question is, is the kernel supposed to put up with
multiple writes to /proc devices and interpret them as one command if
the file is not closed in between? This could become a problem if
/proc files ever get big.
Imaging writing 10K of "ip rules" into a future kernel. No way to do
an atomic write in that case, so you get it chopped however careful you
are.
>From a "principle of least surprise" I would say bash and the kernel
are both borderline here. Bash should buffer a -little- more and the
kernel should be a -little- more tolerant on this policy if possible.
(mostly so that proc can handle big requests, should that even become
necessary.)
--kmb203@psu.edu----------------Debian/GNU--1.3---Linux--2.0.30---
OK, so you're a Ph.D. Just don't touch anything.