[Tinyos-help] Tymo: access with Uart...() on msg returns corrupted values

[Andrey Gursky]

-------------------------
Briefly:
Tymo: after sending msg by Radio - msg in sendDone() cannot be accessed 
with Uart...() if sendDone() failed because of no ack.
-------------------------

Hi Romain!


You know the standard BaseStation application. The messages coming from 
Uart are bridged to radio and handled by Uart...(). If I use Tymo and 
the message couldn't be sent, I can only access it with Radio...() 
calls, but it is incompatible with the standard way and it must be 
additional variable that would show, whether it is the first call to 
send, then use Uart...(), or the next one, then use Radio...().
So I hope this behavior of Tymo could be fixed so it would be maximal 
compatible with the standard approach.


This bug could be not so easy reproduced in TOSSIM-SF. I use some 
applications, that load CPU at 100% and start the simulation under 
Valgrind. Then I send one message throw serial forwarder and wait till 
the routes are deleted. And then if I have "luck", I send the next 
message, that will be delivered but sendDone() says FAILED (FE: The 
message was not acked => FAIL) and just after that in sendDone() I 
access the msg and become the expected values, but just after that in 
the radioSendTask() the values are crazy:

before send:
DEBUG (0): AG: sending message via radio to 1 with addr_from=0, len=2, 
payload=2587, source=0, id=9.
----------
in radioSendTask()
DEBUG (0): AG: sending message via radio to 1 with addr_from=1, len=8, 
payload=513, source=0, id=9.

The output from Valgrind and TOSSIM-SF is attached.

------------------

And one problem I mentioned early with the delays (<30 s, >40 s) relays 
also on the thing, when routes are deleted and the message should be 
sent and it will be delivered, but sendDone() returns failed and then 
it's going in a loop. The messages can't anymore get the ack. And the 
delay between these crazy retries is about 120 ms (just the same as 
while finding new routes, because while the routes are here, the delay 
on motes is about 20-60 s). Could you find out, what is the problem? I'm 
trying to think up something could help you to localize the bug, but 
till now only results described on the top.


Best wishes,
Andrey
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: testTymoLoopValgrind-2008_10_02.txt
Url: https://www.millennium.berkeley.edu/pipermail/tinyos-help/attachments/20081002/a438452f/attachment.txt