Problem with ftn95 v 9.05

The attached program stops in Plato without any failure message. The option “Fail on underflow” is switched off. This behaviour is different from v 8.95. KL

      program test110

         use, intrinsic :: iso_fortran_env
         implicit none

         integer, parameter :: sp  = real32
         integer, parameter :: dp  = real64
!!!      integer, parameter :: sp  = kind (1.0)
!!!      integer, parameter :: dp  = kind (1.0d0)


         real (sp) :: a, b, d
         real (dp) :: c
         logical   :: test

         test = .true.
!!       test = .false.

         if (test) then

            a = 2.89e-3
            b = 8.30544_sp
            c = 95.0428_sp

            d = a * b**5.5 * exp(-c)

            write(*,*) d
            read (*,*)

         else

            a = 2.89d-3
            b = 8.30544_dp
            c = 95.0428_dp
!!          c = 80.0_dp

            d = real (real(a, dp) * real(b, dp)**5.5_dp * exp(-c), sp )

            write(*,*) d
            read (*,*)

         end if

      end program test110

KL

The run does indeed pause at the read(,) statements but this is the correct behaviour and has not changed between versions.

Should'nt it show the result first? Pressing then 'Enter' leaves the program. That is what another compiler is doing. In my environment (Windows 11 Pro latest version) the program simply stops and does nothing.

Yes it shows the result for me and the read is the last statement so it does exit.

I have checked using the v9.05 download.

Paul thank you for for efforts.

The options checkmate x64 and .Net, debug x64 and .NET, release x64 and .Net work for me also. But not the others.

If you don't get the expected result for Win32 when running the app from Plato, try runnning it from a DOS box.

You can open a DOS box from the View menu.

I obtain the same results, when I directly call the executables:

Checkmate ... Debug ... Release ... - NET - Win32 - x64

                            ... run 110.exe

To me it seems that Win32 needs checking. Klaus

Klaus

I have checked and it works fine for me.

Paul, thank you very much for your help.

I have de/reinstalled ftn95 9.00, I have upgraded to 9.0.5, I have used the DOS Box in Plato as well as the Windows PowerShell. None of these changed the reported behavior. This specific problem blocks in my environment all win32 options (which is not the case for gfortran).

My windows version is: Edition Windows 11 Pro Version 24H2 Installiert am ‎08.‎11.‎2024 Betriebssystembuild 26100.2314 Leistung Windows Feature Experience Pack 1000.26100.32.0 Best regards, Klaus

Klaus

Thank you for the feedback. At the moment I can't think of a way forward.

Klaus

Can you describe the faulty behaviour please. Is an exception raised and is any information provided about it?

Virus checker issues.

From time to time we get reports of problems with virus checkers. Virus checkers such as Windows Defender are continuously being updated as new viruses are reported and your virus checker may be updated literally whilst you sleep.

So one of your applications that you could build yesterday will no longer build today.

I guess that the technicians who work on virus checkers look for a key signature (that is characteristic of the virus) within the header of the application. If by chance this signature is in your application then you get a false positive and the application is rejected. The header is built by the linker (in this case SLINK for Win32).

There is not much that we at Silverfrost can do about this apart from providing a warning.

Users can briefly switch off their virus checker to see if the problem is fixed and hence caused by the virus checker. If it is then they can notify the supplier of the virus checker so that they will be able to refine it and avoid the false positive.

One could try running the linker in administrator mode. Otherwise try switching off the virus checker whilst linking.

Or switch to Linux and use virtualization software like WINE to bring you Windows there. Nobody harasses you there with all that mandatory updates and reboots. The only what 'useful' Windows doing last quarter century is just changing flashy screen savers and charging a half of globe fees claiming it became 'faster'. How much we all spent on Windows in our lifetimes?

Silverfrost does not currently provide support for FTN95/ClearWin+/etc. under Linux.

It seems that I got it running under Plato by ... Rebuild Dependency ... Clean ... Rebuild ... Start Run

But I cannot find the corresponding option 'Rebuild Dependency' in the Visual Studio Plugin. Many thank for your help! Klaus

PS. I forgot to describe what happens when the problem occurs. Exactly nothing. There is no information and I can stop the run only via 'Stop Debugging', independently whether I have selected checkmate or release.

Klaus

Another user has reported exactly the same issue. I will try to install 24 H2 on one of my machines and investigate further.

At the moment it seems that the new version of Windows is causing the default exception handler for Win32 fp underflow to fail.

In theory you can provide your own exception handler via TRAP_EXCEPTION@ and FLT_UNDERFLOW but I will investigate first.

The code for the default exception handler is located in salflibc.dll so a previous version of this DLL might provide a temporary fix but better to wait.

Quoted from PaulLaidler Silverfrost does not currently provide support for FTN95/ClearWin+/etc. under Linux.

FTN95 under all numerous virtualization incarnations and all possible distributions of Linux is also working under same ol Windows. And doing everything it do under standard Windows. You also can chose any previous version of Windows you want, including those which have no exceptions problems 😃 As it works same way all do last 30 years it does not need any special support. Best Windows-like look and feel is with Linux Mint. Worst Ubuntu desktop (server is ok)

Dan

That is good news but the fact remains. At the moment we are not offering to fix any Linux issues.

Klaus

I am now able to reproduce the failure under 24 H2. There is currently no known work-around. The default exception handler (for fp underflow) has not changed for many years so I don't understand how it can work for any version (of salflbc.dll).

Please confirm that it works for you under 24 H2 and versions of FTN95 etc before v9.0.

I have sent the details per personal email. Klaus