Improving Windows support

[Alexis S. L. Carvalho]

Thus spake TK Soh:
> On 4/16/07, Alexis S. L. Carvalho <alexis at cecm.usp.br> wrote:
> > Thus spake TK Soh:
> > > Bottom line, we can't have a distributed VCS like Hg without the
> > > capability to merge. So IMO, Hg should provide it's own internal
> > > merging support. For now, perhaps one (written in python?) can be
> > > bundled with the Windows installer. Though some hacking might be
> > > required to make it work 'out of the box'.
> >
> > I've just pushed a few changesets to crew that import the three-way
> > merge code from bzr as contrib/simplemerge.  This is not supposed to be
> > the ultimate merge tool - it's just a simple replacement for systems
> > without a usable /usr/bin/merge.
> >
> > I've also changed the test suite to use this simplemerge by default
> > instead of relying on the presence of a program called merge, which I'm
> > hoping will improve the situation on windows with pysh and maybe fix
> > some problems on Solaris reported by Giorgos Keramidas a while ago.
> 
> Thanks for quick turnaround, Alexis. Now we are I big step forward the
> Ultimate VCS;-)

I had it mostly ready last week, this thread just reminded me to finish
the last few bits.

> BTW, can you point me to the issue Giorgos reported. The last time I
> ran the test suite on Solaris 2.6, everything seemed fine.

http://people.freebsd.org/~keramida/hg-test/crew/2007-03-22.fe0fe0b4d73b/

IIRC, the "Network is unreachable" messages are caused by listing ::1 as
one of the addresses of localhost, without having an IPv6 interface.

And then there's a super merge program that never fails (even when it
should :)

Still regarding Solaris, Thomas mentioned that changeset 949516072388
"broke" test-hook - since older Solaris doesn't have unsetenv, we end up
leaking environment variables from one hook to another.  That changeset
doesn't really change the situation - it just changes the test in a way
that makes it print the leaked variables, too.

Alexis