On Wed, Feb 12, 2014 at 12:16 PM, Andrey Repin wrote: >> The strange thing is that gfortran does compile the code, but >> once compiled, the executables have strange behavior mainly involving >> problems reading in data files. ... > > And this is finally the information, that we can work with. > My wild guess is that your "colleagues" making certain assumptions about > files, that not always true on other systems. > I.e. opening a file in text mode, and then treating [its] data as binary ...
Since the problem occurs going from 32-bit to 64-bit Cygwin, it sounds to me like all-the-world's-a-VAX syndrome. I bet there are places where it reads from files and assumes that if it reads N words into integers, that is N 32-bit quantities, or something like that. I haven't written any Fortran since the 1980s, but I bet there are types that have changed size due to the switch to 64-bit and that results in reading incorrect values from files, including reading some of them from the wrong file offsets, and hitting end-of-file at a different point. Are there any switches to gfortran that control this? I took a brief look, and the only thing I saw was the -finteger-4-integer-8 option and friends, but that sounds like going the opposite direction you want to go, and the -ff2c option that generates C code. You could run the code through g77 on 32-bit, gfortran on 64-bit, and compare the C code generated for the file handling. But I suspect that's more hassle than you want to get into; it doesn't sound like fun to me. Good luck. -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
