[fpc-pascal] Boolean types

[Sven Barth]

Am 04.01.2014 00:17 schrieb "Mark Morgan Lloyd" <
markMLl.fpc-pascal at telemetry.co.uk>:
>>>
>>> b5500centralcontrol.pas(119,18) Error: Incompatible types: got
>>> "ShortInt" expected "ByteBool"
>
> >>
> >> b5500centralcontrol.pas(119,18) Error: Incompatible types: got
> >> "ShortInt" expected "ByteBool"
> >
> > You need to cast nevertheless. It's just that they can be used in
> > place of e.g. BOOL parameters for WinAPI functions whereby the
> > argument value is passed like the underlying non-Boolean datatype.
>
> Thanks Sven, noted and understood. In practice I've changed them all to
qwords since this is the native size of the system being emulated, which
leads to another question. If I have a field and property like
>
> fPB1L: qword;
> ..
> property PB1L: qword read: fPB1L write SetPB1L;
>
> procedure ...SetPB1L(p: qword);
> begin
>   Assert((p = 0) or (p = 1));
>   fPB1L := p
> end;
>

And why don't you use Boolean64? That's a 64 bit Boolean type with True
being 1 and False being 0 (like the normal 1 Byte one).

> is there an unobtrusive way of changing the property to effectively be
>
> property PB1L: qword read fPB1L write fPB1L;
>
> when assertions aren't being generated? The obvious way would be to use
lots of  $IFOPT C+  but is there anything more concise and maintainable?
Can methods be inlined, and could the compiler successfully optimise-out a
method which was a simple assignment?

As long as you don't declare the getter/setter methods as "virtual", but as
"inline" I *think* that it should work. Just test it and look at the
generated assembler code (using -al option). If it works then then there
should be no "call" instruction.

Regards,
Sven
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freepascal.org/pipermail/fpc-pascal/attachments/20140104/67c8c4d2/attachment.html>