[R-SIG-Win] Some suggested changes to Windows makefiles with respect to experimental toolchain

[Avraham Adler]

On Tue, Sep 8, 2015 at 7:11 PM, Duncan Murdoch <murdoch.duncan at gmail.com> wrote:
> I would commit the patch if it made *no* changes to the current
> compilation scheme.  I would not commit it if it made changes that I
> can't judge.


I think I've worked out the necessary changes. I've tested the
changes/patch on Windows 64bit using:

* 4.6.3 toolset (no LTO of course, standard BLAS)
* 4.9.2 toolset (no LTO, standard BLAS)
* 4.9.2 toolset (with Fat LTO objects, OpenBLAS)
* 4.9.2 toolset (with slim LTO objects, OpenBLAS)

and it compiled and passed make check each time. Unfortunately,
though, I do not have access to a 32-bit Windows platform on which to
test compilation. While there are changes to six files, the user will
only see a difference in Mkrules.dist, where there are a couple of new
options. I also tried to explain a bit, and changed the default mtune
from "core2" to "generic" which will work on any version of GCC, and
is recommended when it isn't known what platform will be used for
compilation.

What should the next step be? Should I just post the text of the
result of diff here? Should I e-mail someone a patchfile, the six
revised files, or both? Is this something to list at
<https://bugs.r-project.org/bugzilla3/>?

Thank you,

Avi