On 4/19/2012 8:39 AM, Earnie Boyd wrote:
>> Win64 is getting pretty common now. Perhaps it is time again to
>> consider combining efforts with the mingw-w64 project.
> I would rather consider patches to winsup/mingw and winsup/w32api,
> maybe even creating winsup/w64api.
As you (Earnie) are probably aware, but the rest of this list might not
be, the cygwin folks are already looking at switching to use mingw64's
w*api headers in their tentative planning for cygwin64.
>From that discussion, it was pointed out that there is *NO* reason to
separate "w32api" libs, src, and headers from "w64api" libs, src, and
headers -- at least, at the source level (sure, the .a's will be
different and need to go into different places -- different $build
triple? x86_64-pc-mingw32?). The WinSDK doesn't seem to need to
separate their 64bit and 32bit sources; the mingw64 folks don't seem to
need that; why should we needlessly complicate our implementation, when
we have two independent demonstrations that such is not necessary?
Sadly, I think we need to (restart) the discussion we had previously
about the licensing issues vis. mingw.org vs mingw64's policies.
Something was brought to my attention in that regards; I'll PM the
necessary parties in a few days.
Licensing issues aside, IMO mingw64's w*api headers/libs are superior to
ours. Their runtime library support (e.g. libmingwex, startup objects,
etc), AFAIK, differs only cosmetically from ours, although I'm not sure
about the status of long double support or C99 functions in either of
them. The key bit is organization: their runtime lib support is
co-mingles with their w*api stuff, so it's hard to "just use mingw64's
w*api" without also pulling in their libmingw-related libs and headers.