Re: [lwip-users] With logs ... Re: Delayed Ack causing problems? Where to call tcp_nagle_disable()?

[FreeRTOS Info]

Excellent.  Thanks.  I will update as necessary.

Output message are not always easy to view in embedded environments.
What I should do is run all lwipopts files through the Windows
environment, where viewing the warnings is easy.


Regards,
Richard.

+ http://www.FreeRTOS.org
Designed for Microcontrollers.
More than 7000 downloads per month.




On 13/09/2011 20:54, address@hidden wrote:
> FreeRTOS Info wrote:
>> That would be excellent - and exactly the feedback I would look for.
> 
> Alright, so it turns out the lwipopts.h *was* the reason for the poor
> performance. Using the LPC17xx one, I had the same poor performance with
> my win32 port. I got it working by changing the following line:
> 
> #define TCP_SND_BUF                (TCP_MSS*2) // was: 2048
> 
> That ensures that TCP isn't limited too before sending 2 mss-sized
> segments (which is the borderline for delayed ACKs on most systems).
> 
> BTW: there *are* sanity checks for these defines in init.c. However,
> they are (yet) runtime checks that output a warning via
> LWIP_PLATFORMDIAG in LWIP_DEBUG mode, so there's a high risk of not
> seeing them. Given the time we spent on this, I'll change them to
> compile-time errors that can be disabled. There are also other sanity
> checks that fail with your configuration. While these can be ignored in
> the very special case of the webserver, I'd strongly recommend to change
> your lwipopts.h to obey all of them.
> 
> I'll update init.c in git these days, and I'll have to adapt the default
> values in opt.h, too, since they fail the sanity checks, too :-(  -->
> bad example!
> 
> Simon
> 
> _______________________________________________
> lwip-users mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/lwip-users
>