Hi,

I know the the library change since a long time now, but the problem is not here and I'm sure of that. I don't use the ride's library ( desactived the option in ride) ,and I have MY library in MY project ( the .c and .h files). OK my library is old ( V0.3) but it's ok for me.

When I open and build and debug my project on ride version 7.30.10.0159 with the ride7 patch V7.30.10.0169 and with the rkit arm 1.26.10.0130, the systick interrupt don't work !!! ( a sort of while(1° and I did seveveral test on it.
BUT when I open the SAME PROJECT with the SAME library, without changing anything, but with the ride V7.01.0002 and the rkit arm 1.03.0004, the systick interrupt work PERFECTLY !

For that, I'm sure now that it's not a problem of library !!!

Maybe the startup file, but it is the same !!!

Why I have this problem ??

Thanks,

Hi,

For sure there are people who use the systick interrupt with the latest versions of Ride7, RKit-ARM and ST lib.

How do you observe that "the systick doesn't work"?
Did you set a breakpoint in the ISR?
Do you just observe some memory location that the ISR is supposed to modify?
Did you try the systick example provided with the latest version of the lib?

One of the differences between these versions of the RKit-ARM is the version of the compiler. The new one optimizes a lot better.

So in some cases, like some delay codes or variables shared between main process and ISRs, the compiler will generate very different (but valid) code. The typical mistake is that a variable is not declared volatile when it should be. This was going unnoticed with old versions of gcc but new versions optimize this much better and these coding errors now trigger functional errors.

Typically, things like this (erroneous) code were making a delay before, but not now:

void delay(unsigned long cnt)
   {
   while(cnt--);
   }

The C standard permits the compiler to remove all the code in this function as optimization. The old gcc didn't do it, but the new one does...
It should be rewritten this way:

void delay(unsigned long cnt)
   {
   volatile unsigned long mycnt=cnt;
   while(mycnt)
      mycnt--;
   }

I suggest you make sure that any delay function uses a volatile variable for counting.
And also that any variable that is accessed by both the ISR and the main process (or another ISR) is declared as volatile.

You might also give a try at disabling optimization. If that makes the problem disappear, then you can be sure that it comes from a coding mistake, probably a missing 'volatile'. If the problem persists, the test doesn't prove anything.

Now, if I remember correctly, the old library code itself included some of these wrong codes. (delays optimized out, etc.) So I wouldn't be surprised if the old lib wasn't working with the new compiler at all. I would either use the old lib with the old Ride, or the new version of everything. Don't mix versions too much, as you end up using configurations that nobody else tested before...

Best Regards,

Vincent