[ESS] starting directory not detected correctly anymore

[Vitalie Spinu]

I think this is a consequence of the recent feature which sets the default
directory to the package directory. It looks like it treats your tmp directory
as a package. @Lionel?

  Vitalie

>> On Thu, Nov 30 2017 18:59, Paul Johnson wrote:

> Greetings.

> I want to have ESS know the working directory from shell pwd, I don't
> want it to ask me.

> I have a setting in my init.el that used to work that way:

> ;; start R in current working directory, don't let R ask user:
> (setq ess-ask-for-ess-directory nil)

> Recently, it has stopped working. The symptom of the problem is that
> the R session working directory trims off the last element in the
> path. This is in Ubuntu 17.04 with emacs 24.5.1 and ess 17.11.

> When I open a file from command line, say in

> $ cd ~/tmp/R/
> $ emacs testme.R

> and then I hit the bug blue icon to start R, then getwd() shows
> "~/tmp". The last element in the path is lost.  Same happens if I
> start R with "Alt-x R", so don't hate me for liking your pretty blue R
> button.

> I notice that M-x eshell gets it right, it opens a shell in the /R
> directory (the correct one). Also Emacs File "Open Directory" also
> gets it right. Its just the inferior ESS *R* session that doesn't get
> it right.

> If I remove that line from init.el, then the ESS process stops and
> asks me what directory what I want and it always guesses correctly.

> I never saw this happen before today when opening a pre-existing R
> file from the shell.  I have seem similar in past if I have an Emacs
> session open and close the R session and re-start a new R session
> without closing Emacs.  That second instance almost always has lost
> the "R" from the file path, and shows the parent directory.

> Would somebody try it and tell me if I've just gone all the way off
> the cliff toward crazy?

> Suggestions welcome (except concerning cliffs), thanks as usual.

> pj