[darcs-users] can we remove --no-pristine-tree?

[Trent W. Buck]

Eric Kow <kowey at darcs.net> writes:

> - we /could/ perhaps implement --no-working-directory, which is
>   conceptually cleaner for push-only repositories, in my opinion
>   http://bugs.darcs.net/issue431

+1

I've always been too scared to use --no-pristine-tree in case I forgot
to disable commands, and because the working tree was still there so it
wasn't obvious that I was looking at a "don't make edits!!" repo.  I
could disable commands in _darcs/prefs/defaults, but I would probably
come back in six months and think "which idiot disabled all these
commands?  I'll enable them again!"

I've experienced hg and bzr "no working copy" repositories before, and I
think it's more obvious to spot that you shouldn't be making edits in
them -- because you can't! :-)

BTW, Darcs also wins over hg and bzr in this case, because _darcs is
visible with ls.  Because hg and bzr hide their metadata, I have *often*
become confused about what all these stupid "empty" directories are
doing, and only been saved because I ran "rmdir" instead of "rm -rf".

I think --no-working-tree should also set _darcs/prefs/defaults to
disable commands that operate on the working tree, such as whatsnew and
record.

> - on the other hand, for hashed repositories --no-working-directory
>   would be less transparent than a hypothetical --no-pristine-tree,
>   because of our internal gobbledegooky filenames... we would have
>   to do something like darcs show contents and darcs show files to
>   retrieve files, yuck :-(

This isn't an issue specific to working-treeless repos.  Any tool that
wants to know what *Darcs* thinks is in the repo should be using the
show commands, not just catting whatever happens to be in the working
directory.  Cf. the arguments for/against the --hashed format.