[BusyBox] built-in commands problem

[Rob Landley]

On Friday 24 September 2004 07:16, Doctor Bill wrote:
> It is a mistake to assume /proc will even have been mounted.  I can
> think of several solutions, but all of them involve adding more code
> size than anybody would really want for a busy box solution.   I
> believe normally programming practice is to first try argv[0], cached
> at startup time, then try searching the path, finally try the /proc
> filesystem.
>
> Worse case scenario, a chroot has been done to a read only disk
> without the executable or /proc filesystem.  In that case a user mount
> of /proc on an available directory is about the only solution, but
> most programmers would instead bomb with an error descriptive enough
> so the user would know they needed to somehow mount /proc before
> getting to state next time.
>
> No algorithm is going to work 100% of the time, and one to work
> 99.999% of the time would probably be a library all in itself.  So
> start with the simplest, and if that is not good enough add  in the
> next simplest until it is "good enough".
>
>                                     Bill

Running a built-in command that fails because proc is not accessable from the 
current environment is a totally separate problem from being unable to run 
the command at all due to a chroot orrunning an executable other than 
/bin/busybox in the first place.

In general, a command that barfs because it can't access proc should have some 
variant of a decent error message.  Maybe we need a bb_must_open( ) or some 
such that prints the filename and dies if it can't open it...

That's a totally separate issue from the built-in commands mess.

Rob
-- 
www.linucon.org: Linux Expo and Science Fiction Convention
October 8-10, 2004 in Austin Texas.  (I'm the con chair.)