Description of problem:
Our large Fortran 90 software has recently started failing some of its unit
tests with the following runtime error:
----
Fortran runtime error: Expected INTEGER for item 4 in formatted transfer, got REAL
(a,i8,2f8.2,i8,i8)
   ^
----

An examination of the code shows that the format string and arguments are
correct (and it works fine with other compilers). The error message is also
either inconsistent or incorrect (item 4 looks like the second f8.2 format to
me, while the caret is pointing to the second item). I managed to reduce the
problem down to a simple example, although given that the problem mysteriously
disappeared when seemingly unrelated variables and code sections were removed,
it suggests a bug in the generated code or runtime checks. (Our unit tests also
used to work with the same gcc version, although nearby parts of our code were
changed just before the tests started failing.)


Version-Release number of selected component (if applicable):
gcc-gfortran-4.1.2-33
libgfortran-4.1.2-33

How reproducible:
Always.

Steps to Reproduce:
1. Get test.F90 (attached).
2. % gfortran test.F90
3. % ./a.out
  
Actual results:
Program prints the following output:
---
Fortran runtime error: Expected INTEGER for item 4 in formatted transfer, got REAL
(a,i8,2f8.2,i8,i8)
   ^
?
---
(last line appears to be garbage, but is entered here verbatim as seen on my
terminal.)

Expected results:
Program runs without complaining about the format string, e.g. when compiled
with ifort the following output is seen:
       0    0.00    0.00       0       0
(Note that this is a cut down example, so I don't bother to initialize any of
these variables. So the output may be garbage too. But you shouldn't get the
format string error.)

Additional information:
I guess this an upstream problem, since the same program compiled with gfortran
4.2.2 on an Intel MacBook Pro (OS X Leopard) exhibits the same problem.