Better bounds inference on checked arrays

[john-h-kastner]

int *a[];
void b() { a[0][0]; }

converts to

_Array_ptr<int> a;
void b() { a[0][0]; }

a should be something like _Array_Ptr<int> a _Checked[1]

[mwhicks1]

Could we do something fancy to figure out what the length actually needs to be (1 is the least it could be, but it could be more) ?

[john-h-kastner]

Could we do something fancy to figure out what the length actually needs to be (1 is the least it could be, but it could be more) ?

Yeah, that would be better. In the simplest cast, if we see a[10][10], it makes no sense to have a converted to _Checked[1]. Hopefully the existing array bounds inference code already does most of the hard work here.

[kyleheadley]

The original issue was patched successfully. Later comments discuss Checked array bounds.