Hello Paul,
I waded through the patch and couldn't see anything objectionable. On
applying to mainline, I saw:
/home/pault/Downloads/p11b.diff:1034: trailing whitespace.
/home/pault/Downloads/p11b.diff:1042: trailing whitespace.
/home/pault/Downloads/p11b.diff:1084: trailing whitespace
/home/pault/Downloads/p11b.diff:1104: trailing whitespace.
/home/pault/Downloads/p11b.diff:1117: trailing whitespace.
warning: squelched 1 whitespace error
I removed this. "git diff" has the habit of showing trailing whitespace
in bright red.
git apply --verbose did not enlighten me as to where that last came
from but the trailing whitespace is definitely there.
Apart from these signals, the patch applies cleanly. It also passes
regression testing on a fresh mainline.
Going through the new tests with -Wall, rather than the specific option:
warn_undefined_vars_[1-4].f90 produce the new warning plu a plethora
of unused variable warnings (of course!).
warn_unused_but_set_1.f90 does not produce the new warning
warn_unused_intent_out_1.f90 ditto
warn_unused_read_1.f90 ditto
Is it worth triggering the last three on -Wall?
OK for mainline. Do you have any thoughts about backporting?
Thanks for the review!
I have committed this as
https://gcc.gnu.org/g:9d868e49122b4ce45cffa4d3f47f6ef371e80a24 .
Backporting - this introduces new options, probably not.
Does this improve gfortran's score on the Polyhedron diagnostic tests?
An interesting question - I haven't looked (yet). But at least it will
complain when people use IOSTAT and never look at it :-)
Now, let's see if there is any falloout :-)
Best regards
Thomas
