[sheepdog] [PATCH 4/6] use sd_getopt instead of getopt_long

[MORITA Kazutaka]

At Tue, 06 Nov 2012 15:39:31 +0800,
Liu Yuan wrote:
> 
> On 11/06/2012 03:30 PM, MORITA Kazutaka wrote:
> > I'm not sure whether I understand your suggestion correctly, but IIUC,
> > it means that we create sd_opt_param (including a user-defined handler
> > to parse a key-value string) for each prefix (not <prefix,key> pair).
> > Is it correct?  Then, we cannot auto-generate a help message of
> > key-value pairs, no?
> > 
> 
> > Or you mean that the user should define sd_opt_param for each
> > <prefix,key> pair, and define the handler outside of sd_opt_param for
> > each prefix?
> 
> Ah, I wasn't noticing the structure is <prefix, key> pair.
> 
> maybe handler for each <prefix, key> is okay? like following:
> 
> static struct sd_opt_param write_cache_options[] = {
>         {"object", "size", "<size>",
>          "specify a cache size for object cache", A},
>         {"object", "directio", "<on|off>",
>          "avoid using gateway page cache", B},
>         ....
> };
> 
> Then sd-opt user is freed to use the same handler for handler A and B or
> use dedicated handlers for A and B. In either case, users don't need to
> write prefix parser.

Thanks, I got it, it makes sense.  I'll address it in the next
version.

Kazutaka