Fortran runtime error "fixed" by writing output

Question

I've having trouble with some old code used for research that I would like to compile using the Intel Fortran compiler. In a particular subroutine, I get segmentation faults unless I add in a write statement that just outputs the value of the loop index.

do j=1,ne

SOME STUFF

write(*,*) 'j=', j

end

What could be causing my error such that this write statement would fix my segmentation fault? (Note: j is declared as an integer)

thanks, keely

Can you strip it down to a minimal, self-contained example that still crashes? Otherwise it's going to be tough to diagnose; there is probably undefined behavior somewhere in "SOME STUFF", such that the segfault is staved off by some side effect of the write. Basically your classic "heisenbug".... — Jim Lewis
– Jim Lewis, Commented Aug 26, 2009 at 0:00
I'm having the exact same issue. I can see how it could be due different optimizations being used in the compilation. That makes me think that there is still an issue with the code, even if it's no longer crashing. — swimfar
– swimfar, Commented Oct 15, 2019 at 17:40

user229044 · Accepted Answer · 2012-10-11 00:02:33Z

7

Classic ways of causing this type of error which is 'fixed' by inserting write statements:

walking off the end of an array -- use your compiler to switch on bounds-checking and debugging options to check for this;
disagreement between arguments provided to a sub-program and arguments expected. Again, use your compiler if possible, your eyes otherwise.

Odds are 5-to-1 that one of these is the cause.

edited Oct 11, 2012 at 0:02

user229044♦

241k41 gold badges347 silver badges350 bronze badges

answered Aug 26, 2009 at 11:17

High Performance Mark

78.7k7 gold badges109 silver badges168 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Christoph90 Over a year ago

Why on earth would inserting a write statement fix these runtime errors?

High Performance Mark Over a year ago

It doesn't, it 'fixes' them. The inverted commas are important, used here to indicate that the 'solution' is illusory. The true cause of the error is likely to have been one of the two matters covered.

Christoph90 Over a year ago

I sensed that, however the question remains what's going on under the hood that a write statemant can change the runtime behaviour at all in such a case (apart from writing to the unit).

Collectives™ on Stack Overflow

Fortran runtime error "fixed" by writing output

1 Answer 1

3 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related