Make code conform with Fortran standard

[Leporacanthicus]

The original test "writes" to the argument a of s1, which is a constant value. This isn't valid in Fortran, and will fail if constant values are placed in read-only memory.

The original bug-report wasn't about using a constant argument to a function, but simply that the assign to doesn't work for some cases.

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=22290

This patch fixes the test-code so that it is compliant with Fortran standard, by introducing a variable to assign into.

[Leporacanthicus]

At some point, we probably should bring this back to the gcc folks! :)

[tblah]

Agreed. Writing to a literal constant is non-conforming.

[tarunprabhu]

For both the GCC and gfortran test suites, we do not make any changes to the tests themselves. This originally was borne out of licensing concerns. I am not particularly keen on changing this stance.

We have some end to end tests in Fortran/UnitTests. We could add a test like this there. Fortran/UnitTests/hello is an example.

For this particular test, we should just disable it.

[Leporacanthicus]

For both the GCC and gfortran test suites, we do not make any changes to the tests themselves. This originally was borne out of licensing concerns. I am not particularly keen on changing this stance.

We have some end to end tests in Fortran/UnitTests. We could add a test like this there. Fortran/UnitTests/hello is an example.

For this particular test, we should just disable it.

Ok, I'll add it to the "disable test" list and add something similar to the unit tests.

[Leporacanthicus]

Closing, new PR #135 to replace this one.