:ADDPATCH fortran:
Hello world,
this patch switches to byte-wise handling in Fortran intrinsic masks.
Previously, if a user supplied a kind=1 logical array argument to mask,
we converted it to a kind=4 logical, wasting CPU cycles and memory,
which the user presumably wanted to save in the first place.
This version of the patch has been regression-tested on
i686-pc-linux-gnu. A previous version has been regression-tested
on a big-endian system, powerpc-apple-darwin8, as seen on
http://gcc.gnu.org/ml/fortran/2007-08/msg00288.html
and http://gcc.gnu.org/ml/fortran/2007-08/msg00289.html .
There was one regression, which has been fixed in this version
of the patch.
Right now, I can't get anybody to test this on a big-endian system.
Because this was tested already, and a fix is in, I think the risk
of causing a regression is low.
OK for trunk?
Thomas