works only with WRITE(*,*)

[diedro]

dear all,

I wrote e fortran code that use listes and pointers.

It is:

[bash]DO jp = LISTPART(ii,jj)%current%value IF(.NOT. dlist_move_next(LISTPART(ii,jj))) EXIT CALL KERNEL(rp(:,ip),rp(:,jp),lij,W,dW,xij,tk) r = SQRT( (rp(1,ip)-rp(1,jp))**2.d0 + (rp(2,ip)-rp(2,jp))**2.d0) ENDDO[/bash] it does not work, I get asegmentation fault.

The strange things is that if I use a WRITE in the do cycle like

[bash]WRITE(*,*) R[/bash] it works

could someone explam me why?

is it a unix problem?

thanks a lot to everyone

[mecej4]

When you insert a WRITE statement into the middle of a block of code, and the program behavior changes, it is most often because the code is buggy. A WRITE statement may inhibit certain optimizations, and/or change the way in which the bug manifests itself.

A worse situation is one in which the program seems to work correctly for a long time, and insertion of a WRITE statement breaks the program.

You have to locate the segmentation fault and trace back to the causes of the segmentation fault. Such a task can be sometimes very difficult and time-consuming, especially if the size of the code is appreciable and someone else wrote the code.

I have no idea as to what you mean by "Unix problem". A standard-conforming Fortran program should work on any supported/supporting platform.

[jimdempseyatthecove]

Are/is dlist_move_next and/or KERNEL recursive?

Does the error occur in debug build?

Have you tried adding the diagnostics? (gen-interfaces, uninitialized variables, and subscript error checking)

Jim Dempsey

[diedro]

dear you,

thank you, I found the error. It was a wrong allocation in Kernel subroutine.

a question, what do you mean for diagnostics?

is it a way to compile in order to understand if there are some error?

thanks a lot

[jimdempseyatthecove]

I am not in my Linux environment now, the Windows options of interest are:
(Linux options may have different spelling but same functionality present)

-check:bounds (array subscripts out of bounds testing)
-check:uninit(uninitialized (scalar) variables, arrays are not tested)
-gen-interfaces (generates .MOD files with sub/fun interfaces)
-warn:interfaces (uses .MOD files with sub/fun interfaces and verifies callers)

if compiling fixe format also consider

-warn:truncated_source (fixed souce defaults to 72 columns w/ no warning on truncation)

The correct compiler options should be listed in the documentation.

On Windows

ifort -?

displays the command line options. It may do the same thing on Linux.

Jim Dempsey