thumbs up for the watchdog managed support

+1 for adding managed support for watchdog.

This implies to me that it's the double -> int type inference. What's scares me is that this could be messing up calculations all over my code and I might be missing it.

-Valkyrie-MT

Valkyrie is a digger... Well done!!!!

I think this is a great test case and is easy to reproduce, so I opened a ticket just for this. People can vote for it here:

http://netmf.codeple...m/workitem/1320

Good find Stavros!

-Valkyrie-MT

Hi Valkyrie,

I had it posted on the netduino issue tracker since I didnt know if the problem was on netmf or netduino. The link to that is http://netduino.code....com/workitem/6

Cheers

This looks a bit like this issue:
http://netduino.code....com/workitem/2
Or am I wrong?

Could be yeah.

Didnt know people started uploading issues there. If it proves to be different than the one you posted, I will post it there too. Thanks Stefan

System.Math.Pow seems not to be implemented too

I am also getting strange some results on performing math equations with double. For example:

doubleVar = (double)((3000.0 * someInt) / 16777216.0);

In the above case when the int is 824280, the result should be 1473.9282131195068
However the result in doubleVar is -62.07177734375.

I stopped in debug and I can get the correct result in the immediate

Am I missing something?

Known Errata

• StringBuilder.Replace can cause corruption
• Int64.Parse and Double.Parse throw NotImplementedException
• "Caught Exception" in debug window immediately after boot (cosmetic only)

System.Math.Pow seems not to be implemented too

Awesome news indeed. Chris is AppDomains set to true for this one? "CreateInstanceAndUnwrap" is firing again a 'System.NotImplementedException'.. UPDATED: Checked and turns out AppDomains is set to false. Were there issues that made you keep it to false for the RC3 version too? Thanks

Thanks both for replying so fast (damn its a pity I wasn't stuck with something to ask for help), Chris just ignore us when we keep asking about a release date. Better have all the time the team needs to sort out all the code and make sure those floating points stay where they were supposed to. Stavros

Hi, any new on the 4.2 RC3 "build" based on the 4.2 RTM release? We are eager to work with it, as it seems that most of the problems have been resolved (i.e sockets.connect timeouts, external assembly loading). Fingers crossed but I think Secret Labs should be really close to releasing it. Cheers, Stavros

I finally got my watchdog circuit to work reliably. Whenever the program hangs, it reboots the N+ board. To make this work, I added an external chip whose output drives the N+ reset line with a negative pulse when needed. This circuit needs a periodic pulse to tell it that the software is still running, the absence of a pulse for more than a certain time, will then reset the N+; In my case, I set the timeout interval to about 30 seconds. This gives the N+ software time to do very lengthy tasks without the need to sprinkle pulse generating code everywhere.

Question: Is that any interest in such a shield, or would people just like to make their own? The circuit is not complicated, although the chip is a tiny surface mount one.

Hi Robert, could you give us printout of the schematic you used? I am curious if I can run a quick test to see if it can fit my needs too

Cheers,

Stavros

I am seeing the same thing on both versions. It is true that sometimes an exception will be generated, but not always. Sometimes it just locks up the entire machine, including all threads. To overcome this and potentially other lockups, I am adding an external watchdog timer chip. My plan is to use the MAX6751 which seems to have everything I need. Adding this chip to my design will be next week's goal. It's a shame that the ARM's built in watchdog timer is not supported.

Hi Robert,

This is very interesting, we were looking for something like that. Keep us posted if it can work.

Also for making sure a socket.connect timeouts at some point, you might also want to look at this thread. The idea is to create the socket.connect at a thread and then create a "watchdog" thread that sleeps for some time and after that it kills the socket.connect thread.

I have just tried this on an N+ I have got a System. NotImplemented Error in mscorlib.dll against AppDomain.CurrentDomain.CreateInstanceAndUnwrapp(.....)

I used MFDeploy and discovered AppDomains set to false. What should I do?

Hi Erich, I reckon you are using the 4.2 RC1 version of the micro framework. Read Chris's answer above that they will most probably fix that in the next release of the framework. Hope that helps

Thanks, sounds really promising and will definitively have a look

Turns out AppDomains is false on 4.2 RC1 Does anyone know, how this can be changed? 4.1 is fine too, the prob is that it hangs so many times giving us a headache, while 4.2 in netduino plus is much more stable. Cheers

I guess you should add some outputs to see at which point it gets stuck. Maybe the memory consumption gets bigger as time passes by and the netduino crashes... I think it is a good practice to have a thread "observing" the running processes or the system health so it can for example try to reconnect if the socket connection gets dropped for some reason or kill other threads the seem unresponsive and start them over. Very interested on your project, as we are also trying to gather data from external devices (serial port and analogue ports). Good luck with it, and keep up posted

At last, I finally figured it out!

Here's the code of the target assembly being loaded from the SD card as a little-endian .pe file.

• This assembly only references Microsoft.SPOT.Native and mscorlib, both of which are already loaded in the calling assembly below. This is important because assemblies are explicitly loaded. Any dependency also needs to be dynamically loaded.
• Note the bold string in Print(): it will be in the output at the very end of this post.

Hi Fabien, and all others in the thread

Thanks for the contribution, it could eventually help to a remote update of our software.

Has anyone tested this with the netduino plus?

I am using MF 4.2 RC1 and I am trying your exact code (as all my tries have failed) and I receive the same exception in CreateInstanceAndUnwrap you had.


Cheers

Stavros

Chris do you think that with the 50% patch, will the 4.2 RC1 release more ram in the netduino plus?