On 10 March 2011 20:50, Adam Megacz <[email protected]> wrote: > > Max Bolingbroke <[email protected]> writes: >> My installed GHC is looping when building your TcRnDriver, oddly. > > Yikes, that is bad news. You mean that your pre-installed GHC, without > any of my modifications, is going into a loop? Not good. > > Even though I suspect this isn't my fault, I need to look into it. What > version of GHC are you using to boot (I use 6.12) and on what platform? > I will try to reproduce it. If you've run into this problem and > reported it, I bet 2-3 other people did as well and just gave up.
My initial diagnosis was wrong (I was confused since I was doing a parallel make). My installed GHC (which is now 7.0.2, distributed with the Haskell Platform) appears to loop on your module CoqPass.hs. It is possible that it just takes an extremely long time to build - after all, CoqPass.hs is huge! How long does it typically take in your experience? Last time I tried waiting for ~15 minutes with no luck. I'll leave it running this time and see if I get anywhere. >> I'm going to wait for the HP release to (hopefully) > > Hewlett Packard? Haskell Platform :-) >> I guess that equally this could be done implicitly (which I think is >> what TH does) at the cost of having variables at level n shadow >> variables at level m < n. > > Perhaps, but there are papers showing that > > %% <[ e ]> != <[ %% e ]> > > ... so CSP can't always be implicit. OK, that makes sense. > Moreover, in the context of > *heterogeneous* metaprogramming, using CSP comes with a price, so I'm > not sure it's a great idea for it to happen without an explicit request > from the programmer in the form of some kind of annotation (like %%). What do you mean by the "price"? Runtime cost? Or the fact that CSP (at certain types) forces the embedded language to be a superset of the host language? Regarding comments in your last email: > > 2. I think your tutorial would benefit from some examples showing how > > to actually run sth - your run functions are all currently == > > undefined! > > For good reason! Unlike homogeneous metaprogramming there is no > one-size-fits-all definition of "run" for heterogeneous metaprogramming, > nor can there be. Naturally, there is no "run" that works for all languages, but since it is a tutorial it would be cool to see a "run" for at least one particular language :-) > It's definitely used inside code brackets; if there were a way I could > turn it on at syntactical-depth-greater-than-zero but off everywhere > else I could probably do that. Really? I'm probably being daft but I can't see what would go wrong if you turned off rebindable syntax. Can you give an example? Cheers, Max _______________________________________________ Cvs-ghc mailing list [email protected] http://www.haskell.org/mailman/listinfo/cvs-ghc
