pexpect?

We discussed this to death and came up with:

02:25 < mabshoff> Well, I guess we should get the Cygwin port going again ASAP.
02:25 < wstein> I think that's the best conclusion from this discussion.
02:25 < wstein> But the options are:
02:25 < gfurnish|away> +1
02:25 < wstein> (1) Steal code from Qt.
02:25 < wstein> (2) Hack printf's
02:26 < wstein> (3) Hack dll's
02:26 < wstein> (4) Cygwin+Python+pexpect acting as an xlmrpc server.
02:26 < wstein> (5) virtual machine -- VETO -- it is too big and requires a network :-(
02:27 < mabshoff> (5) can be used as a porting tool
02:27 < wstein> Anything else that I forgot?
02:27 < wstein> Yes.
02:27 < mabshoff> (6) expect?
02:27 < wstein> Yes.
02:27 < mabshoff> (7) SFU
02:27 < wstein> I like (4) a lot as an intermediae.
02:27 < mabshoff> (8) Vista POSIX subsystem
02:27 < wstein> Because what is behind 4 could easily change.
02:28 < mabshoff> on the other side of the XMLRPC?
02:28 < wstein> Yes.
02:28 < mabshoff> Sure
02:28 < wstein> It could be anything.
02:28 < wstein> It just has to provide a function to call a given subprocess with an input, and return
02:29 < wstein> the output.
02:29 < wstein> It could use pexpect+python for some, it could use (2) for some, it could use (1) for some, etc.
02:29 < mabshoff> sure
02:29 < wstein> And to do this would involve some rewriting of devel/sage/sage/interfaces/*
02:29 < wstein>   Maybe.
02:29 < wstein> It might be good.
02:29 < mabshoff> yep
02:30 < wstein> The main thing is to not slow down anything we have right now (e.g., calculus).
02:30 < mabshoff> I think the pexpect situation is something we should concentrate on for the Windows portion of Dev1

Carl Witty suggested looking into Emacs:

[19:23] <cwitty> mabshoff: re: pexpect; on emacs for windows, does "M-x shell" work?  If so, that's another source of ideas.
[19:23] <mabshoff> cwitty: let me try.
[19:23] <mabshoff> But I am not sure emacs is compiled with MSVC or MinGW
[19:24] <mabshoff> cwitty: it does
[19:24] <mabshoff> !!!!!
[19:24] <cwitty> mabshoff: does it matter? (as long as it's not Cygwin?)  I thought MSVC and MinGW used the same libraries.
[19:25] <mabshoff> cwitty: not much, MinGW is "good enough", i.e. it uses the right API