On 06/30/2016 07:40 PM, Peter Korsgaard wrote:
"Andre" == Andre Przywara andre.przywara@arm.com writes:
Hi,
I really don't know. This simple bug has cost me at least two hours yesterday, since it was the rather innocent access to a variable that caused the issue. And if it wouldn't have been for Siarhei to point me in the right direction I'd have spend even more time to find a fix.
Agreed. The toolchain will complain loudly about an overflow of SRAM space, but not about BSS access before DRAM is available.
I see two problems: - this fixes sunxi and possibly leaves other platforms unfixed - when the spl on sunxi grows some more, the bss won't fit and you'll have a problem again, except much bigger this time
I'd rather have this fixed on tiny-printf level with a bit of documentation on why that's done the way it's done.