I initially diagnosed #900 as being due to the stack above main being hundreds of bytes on Metro M4. However, on further investigation, it is actually the stack of main itself being over 600 bytes. This is primarily due to f486ead which added a 512-byte on stack buffer for checking whether the contents of the boot_out.txt file matched what was already written to flash.

I think this could be greatly improved if the code which allocates the char[512] array was moved out of line, similar to what is mentioned in the comment above filesystem_init().

See #903 and #900 for more background.

This calls into question whether #903 really will fix the rosie-ci problems I've heard mentioned.