Re: [lwip-devel] Removing the restriction on number of sockets

[Joel Cunningham]

> On Aug 7, 2017, at 11:36 PM, Simon Goldschmidt <address@hidden> wrote:
> 
> Joel Cunningham wrote:
>> It may be worth noting that there aren’t existing example of “static” versus 
>> “dynamic” implementations
>> of data structures within LwIP.
> 
> I should add: there *is* one example of static allocation: the sockets array. 
> And I don't really like it :-)
> 
> However, given the (dumb) standard that says sockets should be an int (I 
> guess because the first implementers
> used an array index?), this is currently a good solution, given the fact that 
> our memp pools also allocate
> the memory at compile time...
> 
> 

My comment was calling attention to the original proposal which introduced an 
option to control which implementation (static or dynamic) was active. I can’t 
think of a case in LwIP where a data structure had two implementations.  I 
prefer NOT to have to maintain multiple implementations.

The closest thing I can think of is the API message passing (with MPU support) 
for !CORE_LOCKING versus CORE_LOCKING.  Having to maintain both implementations 
(one of which is fairly complicated) has created a significant barrier to 
implementing new APIs that don’t fit within the existing “groups” (netconn, 
netif, etc.) of APIs

Joel