[racket-dev] planet-version follies with old planet packages

On Sun, May 15, 2011 at 7:50 AM, Matthew Flatt <mflatt at cs.utah.edu> wrote:
> At Sat, 14 May 2011 20:58:53 -0700, John Clements wrote:
>> I've been having trouble with an older PLaneT package, and I've finally
>> narrowed it down to a problem with the change in version-numberings.
>>
>> Specifically, I install Jens Soegaard's gzip.plt package (which installs all
>> right), and then I can't use it.  The problem turns out to be that planet won't
>> let me run it, because it thinks that the required & provided versions of
>> racket are
>>
>> #(struct:mz-version 40000 0)
>> #(struct:mz-version 501 1005)
>>
>> ... so I just need to wait for version 400.00 to come out!
>>
>> I see how I can hack around this problem by editing the required-version in the
>> info.ss file, but if I'm correct, this means that no planet packages that
>> precede the version-numbering change can be used.  It seems hard to believe
>> that this wouldn't have been noticed, though, so perhaps I'm wrong...
>
> The version number change happened on February 4, but we decided that
> it was a bad idea, and it was reverted on March 28. So, the version was
> 400 for all releases, and it's still 400.
>
> It's possible that some planet packages were created with a pre-release
> version in between 2/4 and 3/28, in which case those need to be fixed.
> I don't think that's the case for "gzip.plt", though.
>
> It sounds instead like you're using a pre-release version of Racket
> from between 2/4 and 3/28. Is that right? If so, the solution would be
> to upgrade.

One other possible wrinkle here is that planet may be doing its own
computation instead of using the version libraries (as the planet
version checking mechanism pre-dates the version library, I believe),
so there could also be a bug there.

Robby