Re: bug in copy-directory

[Lennart Borgman <lennart.borgman@gmail.com>]

On Fri, Jan 28, 2011 at 5:30 PM, Chong Yidong <cyd@stupidchicken.com> wrote:
> Stefan Monnier <monnier@iro.umontreal.ca> writes:
>
>>>> Actually on 23.2.92.1, copy-directory, called interactively or not copy
>>>> the files of directory A to existing directory B instead of copying
>>>> directory A inside directory B.(as a subdirectory of B).
>>
>>> Hmm, this is a bit problematic.
>> [...]
>>> The trouble is that Lisp callers might depend on the old behavior.
>>
>> Couldn't/shouldn't we distinguish between interactive and
>> non-interactive calls, then?  I.e. have the interactive spec turn the
>> B into (expand-file-name (file-name-nondirectory A)) when B is an
>> existing directory.
>
> Wouldn't that be confusing?  "Copy directory A to directory B" really
> ought to mean the same as "cp -r a b".

Isn't the semantics of "cp" broken + undescribed?

I had a directory x1, but no x2. Doing

  cp -r x1 x2

works as I expect it to, i.e. x1 and x2 are identical.

However after a second

  cp -r x1 x2

there is suddenly a directory x1 inside x2.

I really dislike this kind of context specific semantics that is both
unintuitive and undescribed.