In the baseline compiler we take care to check that the stack with all the operands and locals added onto it won't overflow the stack limit. The stack limit is set above the bottom of the stack by what seems a very large 64kb. The overflow check doesn't take into account that some bytecodes are turned into system or runtime calls. These calls place a number of operands on the stack beyond what it being tested for in the stack limit test of the prologue. It seems that adjusting ESP to take into account the operands and locals in the stack limit test is unnecessary as the overflow area is very generous and we're going to go beyond this limit if we need to handle a special case bytecodes. The stack overflow area also seems very large and is possible a strain to the PLOS (or which ever space it winds up in). It seems the story here is quite murky and we should probably tidy up the overflow checking and the overflow area.