Scanning with in-band headers (docs)

[gerd-moellmann]

This is not a software bug, but maybe the documentation could be improved?

I am using an object format with in-band headers. My scan method is therefore called with client pointers. Here is what I wrote as a comment in the scan method, I hope it explains what I mean:

/* Rough edge alert:

   Assume H is the size of the header, and the block to be scanned
   is [B, L). Because we are using in-band headers, this function is
   called with client addresses, i.e. [B+H, L+H)

   Assume further that the last object in the block is [O, L-H),
   or in client pointers [O+H, L).

   The skip method returns client pointers. When skipping over the last
   object, it returns L. So, the scan is over when the skip method
   returns L, not when we reach the client limit passed to this
   function.

   When we erroneously expect a header following the last object,
   there is enough room for one left, we see that there is none.
   MPS doesn't put a padding object there, maybe because the docs
   say that header-only objects are not supported, but that's just a
   guess. */