Current vs. load directory about racket-mode

[greghendershott]

racket-mode does automatically set current-load-relative-directory to match the source file, to ensure relative requires work. But I'm not sure it ought to automatically set current-directory, as well. Do you think it should? Am I being dense wondering if doing that would cause problems in other cases? e.g. What if you're testing a utility, that works on data in another dir?

---

BTW, also it depends where your program is intended to find "foo.rkt":

1. Whatever the current directory happens to be?

   In that case you may need to change the current working directory. In the racket-mode REPL you can use ,pwd to see what it is, and ,cd to change it.

2. The same directory as the source file?

   In that case the program really ought to be the following, which will work regardless of the current directory.

   #lang racket
   (require racket/runtime-path)
   (define-runtime-path foo.rkt "foo.rkt")
   (file->string foo.rkt)

[greghendershott]

On IRC zenspider reported the curdir is reset after each run. Maybe this is the same issue.

I think this is because current-directory is a Racket parameter, and parameters are per-thread.

So this program:

(current-directory (find-system-path 'temp-dir))
(printf "Main thread c-d: ~a\n" (current-directory))
(define thd
  (thread 
   (lambda ()
     (current-directory (find-system-path 'home-dir))
     (printf "Other thread c-d: ~a\n" (current-directory)))))
(sync thd)
(printf "Main thread c-d: ~a\n" (current-directory))

prints:

Main thread c-d: /tmp
Other thread c-d: /Users/greg/
#<thread:/tmp/foo.rkt:7:3>
Main thread c-d: /tmp

The thing is, the REPL runs your program on its own thread (and custodian). Therefore when that thread changes current-directory, the value isn't retained, just like in the example above.

I need to think about the best way to handle this.

[zenspider]

I suspect this is related. I would expect that the pwd be the same as whatever buffer spawned the process. This would match emacs' other inferior modes.

[greghendershott]

So after mulling this over, and also looking at how DrRacket behaves in this regard, I think the best thing to do is set current-directory (and current-directory-for-user and current-load-relative-directory) to the source file's directory on each Run.

[Note that this doesn't try to do anything about preserving (current-directory) as it may have been typed in the REPL. On the contrary, on the next Run, the directory will be reset back to that for the source file.]

I'll push a commit with that, to close this issue. However if with that change it's still not behaving as you want, please let me know.