Marko Koci=E6 <marko.kocic@...> writes:
>> (On the other hand isn't it perhaps too early to speak about building
>> on a windows host at this time at all?)
> Since windows port is in progress, I think it is good to have
> buildable version so errors could be reported sooner.
For what it's worth, I agree strongly here. A failure to build sbcl
from itself on any platform is probably the most important bug that
can be reported, basically for this reason. (An unexpected failure in
the regression test suite is probably the second most important bug;
some more work needs to be done on the Windows platform support to
allow us to run the test suite, unfortunately...)
For what it's worth, the major missing bits that I see in our Windows
* enough composable external-format support to allow CRLF =3D> #\Newline
translation, to make building sbcl itself more reliable.
* run-program support or its near equivalent, to allow running the
test suite and building more contribs.
* sockets and serve-event support, to allow running under SLIME.
(I'm quite sure that other people have their own desiderata, such as
threads, fastcall support, reliable stack/exception unwinding, and so
on; since I don't have occasion to use Windows, their desiderata might
even get implemented before mine... :-)