Nick Glencross (via RT) wrote:
In r9815 I believe there to be a rare problem with the memory allocation during register colouring.
The code is assuming that a maximum of unit->n_symbols exist, but will step outside this range in rare circumstance. The memory allocation and initially setting the flags is ok, but clearing them stomps memory (I believe).
Ah yep. The out-of-range case is coming from PASM registers, that already have a 'color' aka register number. I've fixed this now along with a bigger change in imcc SynReg handling, which I will ci soon.
Nick
Thanks for investigating that, leo
