Photon startup / deep sleep problem in Develop Branch

Troubleshooting
21

Hi,

yes, I did rebuild the firmware. And sorry, I was a bit unprecise. I did see a difference: I could no longer auto flash with pressing the button :-). But it was not really any faster. I will try again, when I am at home, just to be sure.

22

Hi mdma,

are you sure that the core-common and core-communications libs are needed. I removed them and it still builds fine. That was also my understanding, I read somewhere that they are no longer needed.

23

@Stevie, they are NOT needed. Did you clone the ā€œdevelopā€ branch?

24

Oops, I overlooked the ā€œnotā€ in mdmaā€™s comment. I just wanted to confirm that they are not needed. And yes, I am using the develop branch. Sorry for the confusion.

1 Like
25

So I checked again with a stopwatch (although itā€™s pretty inaccurate by hand). It is actually a bit faster without the auto-flash code. It goes down from about 3 seconds to about 2.5 seconds for me. The part where the LED is switched off (with the above code) is shorter. Thanks!

NB: The startup of the core (joining the Wifi) seems to be more stable with the latest pull from github.

26

Woohoo! Thanks for reporting. That's the culmination of a weekend's work with little sleep! :smile:

1 Like
27

Sounds great. I know that feeling! So thanks, I am glad that I reported it. :thumbsup:

28

@mdma

I have no idea how this is working for you. I have removed the extraneous commands from my batch file so to remove the distracting questions about core-comm-lib and core-communication-lib. I knew they probably had no impact here but they were in my batch so I displayed it here. Probably not a good idea as it clouds the issue.

I see today that the make for the develop branch bombs so I have modified my batch to use the photon_043 branch

Here is a cleaned up batch file I run to reclone the code:

c:
cd\spark
rmdir c:\spark\firmware /s
git clone -b photon_043 https://github.com/spark/firmware.git
cd firmware\modules
pause "Set the device to dfu mode"
make clean all PLATFORM=photon program-dfu
pause "Set the device to dfu mode again"
mkdir \spark\firmware\user\applications\photon
copy c:\mycode\myapp.cpp c:\spark\firmware\user\applications\photon\myapp.cpp
cd\spark\firmware\main
make clean all PLATFORM=photon APP=photon program-dfu

This definitely programs both the system firmware and user firmware but does not return from sleep as it should in 2 seconds.

Just to restate for clarity this is the user code I am running:

/* myapp.cpp */

#include "application.h"

SYSTEM_MODE(MANUAL);

void setup()
{
}

void loop()
{
int i = 0;

while (i < 10)
   {
   pinMode(D5, OUTPUT);
   digitalWrite(D5, HIGH);

   delay(500);

   pinMode(D5, INPUT);
   delay(500);
   i++;
   }
System.sleep(SLEEP_MODE_DEEP, 2);
}

This user code fails with a local compile against the photon_043 branch
It works fine with a local compile against the photon_042 branch
and It works fine with a compile in the cloud.

Any ideas what it could be?

29

@HardWater, the new make file in develop requires that you add ā€œPARTICLE_DEVELOP=1ā€ to the make command line to not get the develop branch warning. Though the code compiles fine against develop, I cannot test it since Iā€™m not at home. :smile:

30

Okay Thanks @peekay123

I believe the reason why I was not able to compile against the develop branch this morning is because the make file now requires a new version of arm-none-eabi-gcc that being 4.9.3 20150529 I had the previous version

I tested the above with the develop branch just now and it works fine

@mdma Iā€™m rather curious.

Is it that:

  1. You made a fix to the develop branch that was not working for me because I was running the previous version of the compiler or
  2. That there was a problem with the older version of the compiler itself no fix needed to the develop branch or
  3. You fixed the develop branch and a new version of the compiler was needed?
1 Like
31

Wow, I never thought it would be a compiler issue. Thanks for reporting that!

I have to say that my only guess is some optimization bug is fixed in the compiler with the latest version. Iā€™ve not done anything proactively to fix it other than the fix for the photon needing a reset, which seemed to also fix the wakeup from sleep issue.

cc: @satishgn / @BDub

1 Like
32

Hi,

I actually did not use the version of the gcc which is now required. I would not even know where to find it for Mac. Can you point me to that one?

I was using arm-none-eabi-gcc 4.9.3 20141119. That worked fine for me. There is a newer version there, but not the one required.

33

@Stevie All I can say is this just gets curiouser and curiouser.

The rev you have now is what I was running first thing this morning when I could not compile

Just search on ā€œarm-none-eabi-gcc 4.9.3 20150529ā€ and you will find the new rev

34

Just did that, cannot find it.The version which is currently on launchpad is 20150609. Maybe the 0529 was replaced with that one?!? I think it is in any case better to search for a version ā€œand newerā€.

35

Definitely. Iā€™m going to add that in just a second.

36

@Stevie
Wow that that makes another odd thing. Here is what I downloaded this morning

gcc-arm-none-eabi-4_9-2015q2-20150609-win32.exe  from launchpad

Once I had it installed and ran

arm-none-eabi-gcc --version

I get the following

arm-none-eabi-gcc (GNU Tools for ARM Embedded Processors) 4.9.3 20150529 (release) [ARM/embedded-4_9-branch revision 224288]
Copyright (C) 2014 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

I had not noticed the filename and the output from --version once installed did not quite match

37

Okay, thatā€™s really interesting, indeedā€¦ Okay, so I can actually use that version. Thanks for the info. But of course mdmaā€™s work to use a newer version as well is also appreciated!

1 Like