forums.silverfrost.com

wahorger · Joined: 13 Oct 2014 Posts: 1257 Location: Morrison, CO, USA

In my code, I had inadvertently place a CALL to a WINAPI function (showwindow), rather than a function reference. I included "USE MSWIN" in the function that referenced the API.

There was no error or warning. The effect on the code was to make it all wonky (only way I can describe it). The callbacks did not work for the %rd it was tied to, and only clicking on the upper right "X" would close the window.

I am wondering if perhaps, somewhere, I have something similar that is making my use of HELP all messed up.

But, it seems like the compiler should find this kind of problem.

PaulLaidler · Posted: Tue Apr 07, 2020 8:15 am Post subject:

Can you provide a short demo program that illustrates the problem?

wahorger · Joined: 13 Oct 2014 Posts: 1257 Location: Morrison, CO, USA

The following code will generate one warning (i_log is never used), no errors.

PaulLaidler · Posted: Tue Apr 07, 2020 2:27 pm Post subject:

A call to ShowWindow(0,SW_HIDE) (with zero as the first argument) might be unfortunate. I don't know what it might do and I would not want to find out by testing it. A NULL value for the HWND might even default to the desktop.

wahorger · Joined: 13 Oct 2014 Posts: 1257 Location: Morrison, CO, USA

Paul, this only shows no compiler error is generated.

I have been working on the "wonky" case, but could not replicate with a short example. I'll try a slight change to see if providing a real handle could help show the issue.

DietmarSiepmann · Joined: 03 Jun 2013 Posts: 279

Paul, wahorger,

I observed this, too, for I use the SALFORD compiler and INTEL 64 bit compiler for the same code base: the INTEL 64 bit compiler creates an error, if using call with respect to a function, whereas the SALFORD compiler does not complain.

Dietmar

PaulLaidler · Posted: Wed Apr 08, 2020 8:12 am Post subject:

Maybe I missed the point here.

If you CALL a function like ShowWindow, that is a FUNCTION rather than a SUBROUTINE then FTN95 does not complain. At run time the returned value is simply discarded.

You will find this is also true for the C++ compiler SCC but here it is clearly accepted in the C++ Standard.

I suspect that the Fortran Standard makes no comment on this. If it does then please let me know.