RE: [ECOS] More on ethernet debugging

[Geoff Patch <grp at cea dot com dot au>]

Hi All,

> No, that's way too long. Does "stepi" also take the same time? A source
> line singlestep may still step over many instructions (although much 
faster
> than that, especially over the network). Try turning on remote debugging 
in
> GDB. ( set remotedebug 1 )

Hmmm...I turned on remote debugging and noticed that the delay is occurring 
in the "$s#73" command, which is the actual gdb step command itself.  I've 
put in some code to strobe a test point inside the state machine that 
builds the packets (__add_char_to_packet() in generic-stubs.c).  What I see 
is that there are 4 commands issued prior to the step ("$m..$X..$m..$X"). 
These all arrive and are processed within a few milliseconds, and then 
there's a *long* delay before the step command arrives out of the LAN 
driver.  Very odd indeed!

I'll have a poke around inside the ethernet code to see if  I can figure 
out what's getting hung up.

Thanks for your help.


Cheers


Geoff

------------------------------
Geoff Patch
Senior Software Engineer
CEA Technologies
Canberra Australia
02-6213 0141