This came up during 2010-03 sprint: it's easy to say "darcs obliterate
-a" which is almost always something you don't want to do.  Shouldn't we
require a full "darcs obliterate --all"? (auto-disambiguation is getopt
may make --a work though, but we can always hack around that)

On Tue, Mar 23, 2010 at 06:43:56 -0700, Simon Michael wrote:
> It doesn't seem worth the surprising inconsistency to me. When is it too 
> easy to type obliterate -a ?

I don't have a good answer to that.  I guess it fact that it was a
one-letter switch just made it seem all too easy to plug in, for
example, by typo.

Note that this does not actually delete all patches, due to Darcs's
current behaviour of only obliterating up to the last tag.  I forget
how this interacts with matchers.

> I might type that along with a --matches if 
> I (think I) know what I'm doing and don't want to be prompted.

But you could still type obliterate --all --matches foo there

> Otherwise, I'm not sure when I'd type it by accident.

> If there's to be a safeguard, I'd prefer an extra "You are about to... 
> are you sure ?" prompt, consistent with revert -a, IIRC.

I'm not generally a fan of confirmation prompts (click through
syndrome), although it may be fair to have them in this rare case.

If we were to decide that this -a protection was a good idea, I
would suggest extending it to unrecord.

But Simon's point about the inconsistency is good.  Maybe it's
not worth the risk of confusing folks.

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9