[fpc-pascal] Changing compiler modes

Michael Van Canneyt michael at freepascal.org
Mon Jul 1 16:30:40 CEST 2019

On Mon, 1 Jul 2019, James Richters wrote:

> I am trying to merge a bunch of my pascal units into a single unit, to avoid circular references.   This is quite a challenge because they are not all compiled with the same mode.
> Some units use {$Mode TP} And I can't compile some of the procedures from
> that unit in other modes because I use features that only work in Turbo
> Pascal mode, like modifying a for loop variable from inside the loop.  I
> also have issues with the curly brace { I can't use it as a literal
> character like MyChar:='{'; because it thinks it's starting comment and
> it's very strict in not allowing nested comments with curly braces like
> this Case RunExt of { #134: {F12}
>      Thing1;}
> #3: {Ctrl C}
>       Thing2;
> End;
> But then I have other procedures that just won't work in Turbo Pascal mode, because I have used  % to specify binary numbers like this:
> My_Binary_Number := %10011001;  and that just give me an error in Turbo Pascal mode.. I could just change them to hex, but there are hundreds of them.
> I'm wondering if there is some way that I can have some procedures compiled with {$Mode FPC} and others with {$Mode TP}
>  I see something about {$ModeSwitch EXCEPTIONS} here: https://wiki.freepascal.org/Compiler_Mode    but I can't get that to work..
>  it doesn't say much about it... If I put that in I still can't change mode for individual procedures.

You can't switch modes between procedure, because a mode implicitly alters the
uses clause, and for this reason can only occur before the uses clause.

Some mode-specific features can be switched on or off on a per-feature
basis, but this is not so for all features, and they are also global (i.e.
can be set only once for a unit).

> Alternatively, is there any way to force it to allow circular unit references?

Only from interface to implementation.

If you have such a strange set of units, I would recommend a rewrite in each
case: it doesn't sound like a 'healthy' code base...


More information about the fpc-pascal mailing list