Re: A few Windows build fixes

[Stefan Monnier]

>> Full support for Cygwin-specific file names can only happen on the C
>> level.  E.g., in w32.c:normalize_filename.  We should probably have a
[...]
> I disagree. The right way to get the cygwin mount points is to run
> 'mount -m' and look at the output. Doing so at the C level properly
> involves calling into the cygwin DLL, which is fragile for a non-cygwin
> application.

That's also my impression.  The way cywin-mount.el does it seems good enough.

>> similar function in ntlib, because lib-src programs will need to use
>> it as well (see the example with emacsclient posted here).

I don't think so.  E.g. emacsclient doesn't need to know about those
things, because it should just pass its arguments untouched to server.el
which can then do whatever it feels like doing with it, including
un-cygwinize file names.

> While cygwin-mount.el is too intrusive, its approach of running mount is
> correct. Perhaps a conmbination of the mount table from cygwin-mount.el
> integrated with some tramp support would do it ?

No, the whole point of my discussion is that it should *not* be
a file-name-handler (i.e. unlike Tramp and cygwin-mount) because that's
too intrusive.  If you want cygwin-name support within Emacs (i.e. as
a file-name-handler), then cygwin-mount is the right way to do it.


        Stefan