Re: [lwip-users] [PATCH 1/7] lwip: config sanity check for NETIP

[Andrew Dennison]

On 9/21/07, Goldschmidt Simon <address@hidden> wrote:
> [re: config sanity check for NETIP]
> > +#if (LWIP_NETIF_API && !LWIP_TCP)
> > +  #error "NETIF API is for TCP"
> > +#endif
>
> Why would LWIP_NETIF_API depend on LWIP_TCP? Don't get confused with
> tcpip.c vs. tcp.c! Tcpip.c mainly contains a multithreading-wrapper
> for the lwIP core and has nothing to do with the TCP protocol
> implementation.

I may have the wrong config variable there. This was early on in
getting stuff to compile with various options. Part of the motivation
to post my local patches was to check I hadn't done anything stupid :)

>
> [re: report out of memory by pool name indebug:]
> If that works, it's a good idea for debugging, I think.

Works for me. I added this to confirm which pool was out of memory
before I changed my config.

>
> [re: fix build error with LWIP_UDP=0]
> I'd rather rely on a check in init.c to solve this.

There are a few other defaults that are done like this, so this seemed
like a good idea. This just makes the default sensible if lwipopts.h
disabled UDP, so lwipopts can be relatively small. A check in init.c
would be useful for the case when the user explicitly enabled this
option.

>
> [re: fix warning: 'packed' attributeignored]
> That's strange, I think. I've never seen a compiler warn about this.
> But then again, I haven't used gcc4 so far. Can't you simply turn that
> warning off? I'm using gcc3 and there are some warnings I can't use
> When compiling the lwip code...

Probably, but I prefer to fix warnings rather than turn them off. From
my reading the packed attribute always did nothing but gcc4 now warns.

Andrew