[Toybox] [patch] add u?mount

[Rob Landley]

On 08/18/2013 07:56:29 PM, Strake wrote:
> On 18/08/2013, Isaac <ibid.ag at gmail.com> wrote:
> > On Sun, Aug 18, 2013 at 09:38:42AM -0500, Strake wrote:
> >> From 999f64b2615cd53504e4ad312f7a72eb2170da5f Mon Sep 17 00:00:00  
> 2001
> >> From: Strake <strake888 at gmail.com>
> >> Date: Sun, 18 Aug 2013 09:29:51 -0500
> >> Subject: [PATCH] add u?mount
> >>
> >
> > Good start, but I'd suggest looking at Rob's mount.c
> 
> Damn it; if I had known I wouldn't have written another. I just
> checked the source tree.

It's one I also wrote the busybox version of, spent months getting it  
right last time, and thus have a fairly elaborate set of criteria for  
what counts as "correct".

Last time I sat down to bang on this, the sequencing required doing  
losetup so I could do umount so I could do mount. The losetup saga is  
here, and basically ate christmas break last year:

   http://landley.net/notes-2012.html#28-10-2012
   http://landley.net/notes-2012.html#03-12-2012
   http://landley.net/notes-2012.html#24-12-2012
   http://landley.net/notes-2012.html#25-12-2012
   http://landley.net/notes-2012.html#26-12-2012
   http://landley.net/notes-2012.html#27-12-2012
   http://landley.net/notes-2012.html#30-12-2012

The reason I haven't finished these is people keep sending me new  
commands, and getting those merged and cleaned up is higher priority  
for me than writing new code I already know how to do. (I.E. constant  
stream of interrupts starving base tasks for scheduler time.)

The real priority ones are where somebody sends me a command  
implementation with a design I don't want to use. Last month I got a  
ping that shelled out to a separate ping6 binary for ipv6 support: this  
means I need to write "ping". The "install" command recently posted is  
a wrapper around cp; it doesn't call cp_main() but instead does an exec  
with $PATH search. Since we've _got_ cp, it seems like we should be  
able to just use it without needing to exec ourselves...

But saying "no" to a working command is... unpleasant. So I need to go  
write what I think the command should look like and get that in so they  
can use it...

Working on it. Sorry it's so slow, but I want to get it _right_ or  
there's no point in doing yet another one when there's a dozen  
implementations already out there in gnu and busybox and the various  
BSDs and sbase and sash and klibc and the ancient AT&T stuff SCO open  
sourced before going crazy. I want this version to be _better_ than  
those...

Rob