Re: Gnus UU support

[Eric Swenson <ejs@borland.com>]

Thanks, Andrew for your explanation.  It makes sense to me.  Now, the
more important question, should the Gnus sources (and indeed any Gnu
Emacs sources) try to take this kind of stuff into account?  Should I
attempt to get the Gnus folks to incorporate my change into the Gnus
sources?  I'm kind of leary about trying given that most folks who
write code for Gnu Emacs look down rather loudly at NT-specific
changes.  Folks who run GNU Emacs on non-win32 platforms are
all-to-often apt to say "that's what you get for running on Win32".  

My change just adds a unwind-protected call to cd around the code that
starts the process.  Should I attempt to get the Gnus folks to make
the change or is there another avenue (like you and the other NT Emacs
maintainers)?  

Thanks.  -- Eric

Andrew Innes <andrewi@harlequin.co.uk> writes:

> Not quite - you have tripped over a "feature" of DOS/NT command shells,
> which derives from the DOS drive/directory API.  Each drive maintains
> its own concept of working directory (per process on NT), and the wd for
> a drive can be changed without changing the cwd for the process (except
> that the cwd will be changed by changing the wd of the current drive).
> 
> [Note that on NT, I gather the command shell goes to some trouble to
> make things appear to work this way, because the underlying API no
> longer has separate functions to change drive and wd, although it still
> supports the ability to change drive, setting cwd based on the wd for
> the new drive.]
> 
> So the `cd' command only tells the shell to change the wd of a drive -
> unless you are changing the wd for the current drive, this does not
> change the cwd for the shell process.
> 
> There is a separate command to change the current drive - just the new
> drive letter and colon alone on the command line.  When you select a new
> drive, that drive's wd becomes the process cwd.
> 
> Unfortunately, this kind of solution will be required on NT, until such
> time as there is a decent shell for NT which can be made to behave like
> Unix shells and is commonly available.
> 
> Given the legacy of the DOS command shell, and the length of time it has
> taken Microsoft to make even modest improvements in it, we cannot really
> expect the standard shell to ever have this ability.