OpenEuphoria

2. Re: Rob ??

• Posted by Robert Craig <rds at ATTCANADA.NET> Aug 06, 1999
• 421 views
• Last edited Aug 07, 1999

Bernie Ryan writes:
> I still do not understand the passing a "C" structure in
> Euphoria. You say that 4 or fewer bytes will be assigned
> 4 bytes in a structure.
> Does that mean that a single byte takes 4 bytes ?

A parameter of 4 or fewer bytes will be pushed as 4 bytes
*on the call stack*.

In a C *structure*, you could have 4 single-byte fields declared
in a row. A C compiler will allocate them as one byte each.
Similarly, if you declare 2 consecutive C short ints
(2 bytes each), the C compiler will allocate them consecutively
as 2 bytes each. The rule, as I understand it, is that 4-byte
quantities must start on a nice 4-byte boundary in memory.
2-byte quantities must start on a 2-byte boundary (i.e. an
even address), and single-byte quantities can start anywhere.
The C compiler will leave empty filler bytes to achieve
this alignment.

> How do you poke and peek floats and doubles
> into memory/structure ?
> How much do you allocate for floats and doubles in a structure?

You use atom_to_float64() or atom_to_float32() to create
a sequence of 4 or 8 bytes. You then poke that into memory.
Or, you peek 4 or 8 bytes into a sequence, and use
float32_to_atom() or float64_to_atom().

I believe that both floats (4 bytes) and doubles (8 bytes)
must be aligned on a 4-byte boundary in a standard
WIN32 C structure.

> Normally when I pass the address of a structure to windows
> it knows the layout of that structure's offsets and size because
> of the definition of that structure in the header file.
> When you pass that structure in your format to windows,
> Isn't windows going to be using the wrong offsets when it
> trys to access that structure?

That's the idea. You have to mimic the layout in memory
that a C compiler would use when it compiles a structure
declaration.

About 90% of the fields in C structures under WIN32 are
either 4-byte integers or 4-byte pointers, so it's usually
pretty easy to figure out the offsets of the fields in memory.

If you program in C you won't have to calculate these offsets,
but then you're stuck using C with all of its shortcomings
compared to Euphoria. People who use Win32Lib or Llama
or some other "wrapper" for WIN32 can avoid most of this
low-level byte-counting work and just call the Euphoria
routines that someone else has provided.

Regards,
     Rob Craig
     Rapid Deployment Software
     http://members.aol.com/FilesEu/