[racket-dev] Planet require gone wild

From: Greg Hendershott (greghendershott at gmail.com)
Date: Fri Aug 2 16:19:16 EDT 2013

On Fri, Aug 2, 2013 at 1:25 PM, Matthew Flatt <mflatt at cs.utah.edu> wrote:
> At Thu, 1 Aug 2013 21:55:37 -0400, Greg Hendershott wrote:
>> I'm running HEAD = c126a8a from about 1 week ago.
>>
>> I wanted to try the PEG Planet 1 package.
>>
>> My source file was simply this:
>>
>>     #lang racket
>>     (require (planet kazzmir/peg:2:0/peg))
>>
>> 1. Using command-line Racket, I got:
>>
>> Welcome to Racket v5.90.0.3.
>> ; uncaught exception:
>> '#&"/Users/greg/src/scheme/collects/markdown/markdown/peg.rkt"
>>
>> Huh.
>
> I haven't been able to replicate this. It looks like you're using
> `xrepl', but maybe I'm missing something else about your setup. Can you
> tell me more about how you ran the program?

Well that's a very perceptive question. I ran it using a Racket mode
for Emacs that I cobbled together back when ,en and enter! were having
trouble reloading modules, and when Quack had an issue with Emacs 24.
One thing led to another and I ended up with an Emacs mode, which I've
simply kept using out of inertia. It's on GitHub (not that I've been
seeking for anyone else to use it, just so I can access it):

https://github.com/greghendershott/racket-mode

The part you're asking about is this:

https://github.com/greghendershott/racket-mode/blob/master/sandbox.rkt

I'll wait while you and e.g. Eli wince so hard that you sprain a
muscle ... go ahead. This is probably wrong on multiple levels. Sorry!
Anyway, taking that out of the picture:

>> I had the bright idea of trying this again, in DrRacket 5.90.0.3.
>>
>> 2. Right away, I got a red message:
>>
>> make-directory: forbidden (write) access to
>> /Users/greg/Library/Racket/planet/300/5.90.0.3
>>
>> Strange...
>
> That looks like the background syntax colorer complaining that it's not
> allowed to install PLaneT packages. It's not a nice error message, but
> it reflects something working right (and, of course, the new package
> system doesn't provoke the same sort of error).

Granted, I don't require Planet 1 packages frequently, and the last
time was many months ago. But I'm pretty sure I don't recall seeing an
error, much less one about a write permission (which seems wrong, I'm
pretty sure it's write-able). This message is largely what made me
think something was wonky with 5.90.0.3.

Posted on the dev mailing list.