Re: [lwip-devel] Proposing change to netif state member

[Bill Auerbach]

>-----Original Message-----
>From: address@hidden
>[mailto:address@hidden On
>Behalf Of Sprow
>Sent: Monday, March 19, 2012 2:37 PM
>To: 'lwip-devel'
>Subject: Re: [lwip-devel] Proposing change to netif state member
>
>In article <address@hidden>,
>   Bill Auerbach <address@hidden> wrote:
>> It would be nice to override the void type in void *state in struct
>> netif.h.
>> This is the pointer to be used by the Ethernet hardware port or other
>> user-code and it would be nice to use it with the correct type instead
>of
>> casting whenever it's used.
>
>I just use
>
>  err_t some_netif_function(struct netif *netif)
>  {
>    ethernetif_t *netstate = netif->state;
>    :
>    netstate->mything = 1;
>    netstate->stuff++;
>    :
>    return ERR_OK;
>  }

This is fine, yes.  You will still need a cast here in a C++ environment.
You're right on the code savings.  Where I was casting netif->state is where
I have 4 or 5 one-liners to extract something simple from the low-level
driver (packet counts, link state, connect speed, etc).  I suppose even with
your example a decent compiler will optimize the reference to netstate away.
You do gain seeing into the state pointer in a debugger and seeing the type
it points to.

>> #define NETIF_STATE_TYPE struct myPortDescriptor *
>
>Please at least leave the '*' with the state!

True, I do in fact agree but does mean you can't use a typedef'ed pointer.
Personally I don't care for them - it does seem to be common in Windows API
stuff.

Bill