gfortran and valgrind memcheck

[Tobias Burnus]

Hi Jonathan,

Jonathan Hogg wrote:
> Just wondering if gfortran programs are supposed to be free of bugs
> detected by valgrind --memcheck?
Yes, the goal is that the gfortran library and syntactically
leaking-free, valid programs don't leak memory - and especially don't
use not-initialized memory etc. (Note that sometimes valgrind gives
false errors, e.g. for REAL(10) variables where not all bits are
initialized, but only those which are are used.)

>From time to time we run valgrind on the test suite to find (and then
fix) errors, but this does not mean we found all. Therefore, we are
interested in small, complete programs which show those problems.

> Using the 4.3.1 release compiled from source, most of the bugs seem to
> centre around file functions (and I believe its not my data causing it
> but I could be wrong?)
It is difficult to judge without a complete program to test. I tried to
come up with a valgrind-failing program, but I couldn't. It indeed looks
like a gfortran problem but it still might turn out that it is a bug in
your program.

Tobias