[Harbour] -z switch seems to be overly anti-lazy

[Przemysław Czerpak]

On Wed, 03 Feb 2010, smu johnson wrote:

Hi,

> I will try to take a video of the behaviour and figure it out.  I swear to
> you, that I am not making this up.  If it helps, we are using Clipper 5.2e
> and Blinker.  I will take screenshots of the behavioru to prove what I am
> saying is the case, and hopefully, with your Clipper knowledge :)  we can
> figure it out. Thanks for replying !

I also have CL5.2e and BLINKER and before I replayed I had verified
the results. I do not need any video, screen shots or antyhing like
that. I need only code example. The one you sent works in the same
way in CL5.2e and Harbour (after fixing typo which breaks compilation
so I guess it was not real code but you typed it by hand for that
message only). If you try to create real .prg code example and compile
it using Clipper and Harbour (linker is unimportant here, you can use
also RTLINK, i.e. with cl.bat) then you should find the exact reason
of problem. If it will not be enough for you to resolve it then please
send the code example here with descrption about the difference and
I'll try to help you. Such example should be as small as possible,
i.e. this code illustrates in 1-st line the the at compile time
Clipper and Harbour gives _EXACTLY_ the same results when used with
and without -z switch and the 2-nd line shows that Clipper does
not optimize expressions in macrocompiler so code generated by
macrocompiler is different then code generated by compiler. In
Harbour it's the same.

   ? eval({||qout("hello"),.f.}) .or. .t.
   ? &('eval({||qout("WORLD"),.f.}) .or. .t.')

best regards,
Przemek