OS_Heap 2 / OS_Module 6 differences

The following invalid claim allocates a 32 byte block (presumably a minimum size):

SYS "OS_Module",6,,,0 TO ,,B%

The equivalent OS_Heap call returns the error “Not enough memory (in heap)”:

SYS "OS_ReadDynamicArea",1 TO RMA%
SYS "OS_Heap",2,RMA%,,0 TO ,,B%

So we get two “features” for the price of one. OS_Module should fail the claim as its invalid and OS_Heap should return a more appropriate error.

OS_Heap’s response to a zero-sized allocation has always been silly, especially as it otherwise rounds up allocations to 8n+4 – zero ought to result in a 4 byte allocation really.

OS_Module,6 however rounds up the allocation to a size of 16n+12 before calling OS_Heap, which means it’s happy to do an allocation as small as -3.

As neither of these routines return the size they have actually allocated and there’s no call to read the size of an OS_Module block, one has always had to rely on two assumptions:

• OS_Module uses OS_Heap and exclusively uses DynamicArea 1 for allocations
• OS_Module blocks are always preceded by the block length+4

Also note that because of this undocumented rounding up, OS_Heap,4/OS_Module,13 behave very strangely: If they are used to shrink a block enough, then the block will be freed (and a pointer of -1 returned). However, enough is rather difficult to predict. I invite the reader to experiment and express surprise.