Invalid command in system_get_os_print

[blubb]

Hi,

sometimes I observe a strange fatal exception. I cannot reproduce it consistently, but if it happens, it happens on the restart after an OTA upgrade.

On system start I do this:

struct rst_info *rst_info = system_get_rst_info();
os_printf("reason=%d exccause=%d epc1=0x%X epc2=0x%X epc3=0x%X excvaddr=0x%X depc=0x%X\n", rst_info->reason, rst_info->exccause, rst_info->epc1, rst_info->epc2, rst_info->epc3, rst_info->excvaddr, rst_info->depc);

Two examples of what I see when an exception happens:

reason=2 exccause=0 epc1=0x4021ED6C epc2=0x0 epc3=0x0 excvaddr=0x33 depc=0x0

reason=2 exccause=0 epc1=0x4029ED6C epc2=0x0 epc3=0x0 excvaddr=0x0 depc=0x0

According to the map file epc1 in both cases is system_get_os_print in the NEW image (so the crash does NOT happen while still the old image is active). What is system_get_os_print? What does it do? When is it called? I do not use this function.

My OTA update is done using spi_flash_erase_sector and spi_flash_write in the right places. After that I do

system_upgrade_flag_set(UPGRADE_FLAG_FINISH);
system_upgrade_reboot();

Updating works, i.e. the new image runs perfectly (despite the exception message). If I reboot the module after that, all is well, too (the exception does not happen again).

I am using the latest SDK from
https://github.com/espressif/ESP8266_NONOS_SDK

Ideas?

[Her Mary]

The only thing I know is that you should "make clean" after compiling user1.bin, and then to compile user2.bin.
Or maybe you can dump flash to check if the new firmware is downloaded correctly.

[blubb]

I will download and compare the binary next time the problem occurs. It is so rare, it did not occur since I wrote my first post.

But I doubt it has to do with binary corruption. There are checksums and also the exception always happens in system_get_os_print and nowhere else. My guess is that there is a problem right after switching from user1 to user2 (or vice versa) but before the new firmware runs.

Firmware update is done via a HTTP connection and I call

system_upgrade_flag_set(UPGRADE_FLAG_FINISH);
system_upgrade_reboot();

using a timer (1 second) so that the HTTP client gets a proper response and no timeout on a successful update. I noticed by varying the time the problem occurs more often when it is shorter.

[Andrewgvks]

Be sure to post pics when you get it

What amp / effects are you running it through?