PLaneT and automatic compilation bad interaction (was Re: [plt-dev] Re: problem with optimistic compilation)

On Tue, Aug 18, 2009 at 9:48 PM, Robby
Findler<robby at eecs.northwestern.edu> wrote:
> Actually, that can't be the whole story, since it is trying to compile
> errortrace .zo files, not regular ones and that is controlled via
> parameters. So there must be some sharing going on, somehow. Hm.

Oh, I figured that one out; of course, it was the scheme/base
parameters (current-compile and use-compiled-file-paths in this case)
that were being shared and the one in compiler/cm
(manager-skip-file-handler) that wasn't.

At this point, it seemed to me that it would be hopeless to try to get
the original values of the scheme/base parameters (specifically
current-compile), so that the right path would be to copy the
manager-skip-file-handler's value to the new version of that parameter
in the new namespace.

So, I propagated the value of manager-skip-file-handler to the new
parameter in the new namespace in the code in planet/resolver.ss. That
didn't work, because setup-plt is also creating a new namespace (in
plt-single-installer.ss). So, I propagated the value there, too. But
this also didn't work and I got lost at that point (I did see that
there is a namespace creation that happen in setup/unpack.ss and that
code runs while planet installation is happening, but I don't think
that that namespace is relevant here.)

Robby