_Itype_for_any(T) parameter itype duplicated

[mwhicks1]

typedef unsigned long size_t;

extern void *memset(void * dest : byte_count(n),
             int c,
             size_t n) : bounds(dest, (_Array_ptr<char>)dest + n);

_Itype_for_any(T) void
vsf_sysutil_memclr(void* p_dest : itype(_Array_ptr<T>) byte_count(size), unsigned int size)
{
  memset(p_dest, '\0', size);
}

rewrites to

typedef unsigned long size_t;

extern void *memset(void * dest : byte_count(n),
             int c,
             size_t n) : bounds(dest, (_Array_ptr<char>)dest + n);

void vsf_sysutil_memclr(T * p_dest : itype(_Array_ptr<T>) : byte_count(size), unsigned int size)
{
  memset(p_dest, '\0', size);
}

Two problems:

• The _Itype_for_any(T) part got dropped.
• The type of p_dest was changed from void * to T *

[mwhicks1]

Repeating @jackastner comment from merge: This merge makes things better, but there's still an issue when adding a checked pointer into an _Itype_for_any function.

_Itype_for_any(T) void baz(void* p_dest : itype(_Array_ptr<T>) byte_count(size), unsigned int size, int *a) {
  memset(p_dest, '\0', size);
}

converts to

void baz(void *p_dest : byte_count(size), unsigned int size, _Ptr<int> a) {
  memset(p_dest, '\0', size);
}

We'll need to start storing more information about the polymorphic types on functions declarations so we can reconstruct them when rewriting.

[kyleheadley]

This works for me, but I'm leaving it to ask if we have regression tests for this. Also, the parameter itype is rewritten, but we're nearly complete with a major update to that, so additional bugs should probably be posted elsewhere.

[kyleheadley]

As of today, the rewriting is:

_Itype_for_any(T) void baz(void* p_dest : itype(_Array_ptr<T>) byte_count(size) : itype(_Array_ptr<T>) byte_count(size), unsigned int size, _Ptr<int> a) {
  memset(p_dest, '\0', size);
}

duplicating the itype. This is the same of the upcoming generics PR as well.