- Sun Aug 21, 2016 8:28 pm
#53533
OK. I've given up!
Delving into the SD and SdFat libraries to try to find the problems, I've been adding Serial.print statements for debugging, using a callback function in the sketch to get that to work within the libraries.
As I've added these statements, the non-performance has got steadily worse until finally the execution crashes and the sketch gets woken up by a watchdog timer, only to repeat.
I've traced the crash to File::File(SdFile f, const char *n) in File.cpp and by inserting return; at various places, have found that the crash occurs deeper in a malloc() or memcpy() call within that function.
Interestingly, using the String() family of operations, seems to use the heap as well as malloc().
Although the freeRAM() function I submitted here earlier is reporting heaps(:->) of heap available, thus perhaps ruling out a memory leak, my gut feeling is that somewhere the heap management system is running out of some resource - probably related to the number of calls on its services rather than the amount of heap allocated.
Anyway, I'm transferring my development to a Raspberry Pi zero for now although the PCB I've prototyped and the MCP2301x I2C I/O expander library will probably be useful elsewhere.
'bye for now.
Peter.
Delving into the SD and SdFat libraries to try to find the problems, I've been adding Serial.print statements for debugging, using a callback function in the sketch to get that to work within the libraries.
As I've added these statements, the non-performance has got steadily worse until finally the execution crashes and the sketch gets woken up by a watchdog timer, only to repeat.
I've traced the crash to File::File(SdFile f, const char *n) in File.cpp and by inserting return; at various places, have found that the crash occurs deeper in a malloc() or memcpy() call within that function.
Interestingly, using the String() family of operations, seems to use the heap as well as malloc().
Although the freeRAM() function I submitted here earlier is reporting heaps(:->) of heap available, thus perhaps ruling out a memory leak, my gut feeling is that somewhere the heap management system is running out of some resource - probably related to the number of calls on its services rather than the amount of heap allocated.
Anyway, I'm transferring my development to a Raspberry Pi zero for now although the PCB I've prototyped and the MCP2301x I2C I/O expander library will probably be useful elsewhere.
'bye for now.
Peter.
