[BusyBox] [patch] new applet mountpoint

[Bernhard Fischer]

On Fri, Aug 19, 2005 at 04:30:02PM +0200, Rainer Weikusat wrote:
>Bernhard Fischer <rep.nop at aon.at> writes:
>> On Fri, Aug 19, 2005 at 11:40:57AM +0200, Rainer Weikusat wrote:
>>>Given that busybox may eventually include an operating system kernel
>>>and a C library of its own :->, but will certainly never
>>>(substantially) work with a different kernel than Linux, why not
>>>
>>>mountpoint()
>>>{
>>>        cut -d' ' -f2 /proc/mounts | grep -q "$1"
>>>}
>>>
>>>plus a set of similar small shell functions that print the other
>>>information, insofar requested?
>>
>> Well, so let's zap head and tail too?
>
>Well, there may be reasons why someone would want to not have tail,
>but that's not the same question as "Why mountpoint.c", especially
>considering that your implementation is both broken due to various
>races (the stat/ lstat calls, for instance), needlessly

Hm. the sequence is stat /dir && stat /dir/.. || exit failure;
Please explain that race?

>inefficient (pointless memory allocations) and does some stuff which

If you refer to the call to bb_xasprintf, perusing this was smaller than
doing it by hand. Care to explain how you would accomplish this more
efficiently?
What other inefficiencies (apart from the strerror) do you mean? Just
curious.

>is hardly useful (like printing a single \n with printf
>(putchar, anyone?) when told to be quiet) and does some things which

You're right on the putchar. I agree that it looks a bit odd, but
sysvinit does print that newline so i added it to be compatible.

>are just weird (why the 'stat/lstat'-combination)?

My copy of the mountpoint manpage reads:
[...]
Symbolic  links are not followed, except when the -x option is used.
[...]