The problem is the `lazy-require' that commit 949d12e2c6 says "does not seem to cause [...] trouble".
I'm not sure what changed to make the suspicious `lazy-require' problematic after all, but I've changed `lazy-require' to fix it. It should be ok to use `lazy-require' in `begin-for-syntax' from now on. The root of the problem is that `define-module-path' and friends do not work right at phase levels other than 0. I haven't figured out a general solution to that problem, but I have a general workaround (which is now in the documentation): put the use of `define-module-path' in a submodule, and then the submodule can be imported at any phase level. The expansion of `lazy-require' now does that. Along the way, I also fixed bugs related to `require' inside `begin-for-syntax' and submodules with `raco exe'. Also, `lazy-require' now creates module dependencies more correctly (e.g., if a lazily-required module is triggered at compilation time for some other module), and that improvement needed an addition to `compiler/cm-accomplice'. A workaround for v5.3.1 is to add (require (for-syntax racket/match/parse)) to your program. You might also be able to adjust some uses of `match' to avoid the problem, because only certain patterns (such as `pregexp' patterns) expose the bug. At Fri, 09 Nov 2012 21:23:14 -0500, Neil Van Dyke wrote: > My app, built using "raco exe" and "raco distribute", works in 5.3 but > not 5.3.1. > > With a clean build under 5.3.1, invoking the app binary gives the error: > > standard-module-name-resolver: collection not found > collection: "racket/match" > in collection directories: > /home/user/racket/rackout/dist-tree/lib/plt/rackout/collects > context...: > standard-module-name-resolver > #%embedded:g6830:mime-types: [running body] > #%embedded:g6537:servlet-env: [traversing imports] > #%embedded:g6514:serve: [traversing imports] > #%embedded:g5856:dispatch: [traversing imports] > #%mzc:rackout: [traversing imports] > loop > > However, clean build under 5.3 works fine. > > Running the code as "racket -t myfile.rkt" works fine in both versions. > > Or, rather than cause being in raco exe/distribute, I guess it could be > in some trickery of "web-server". > > Neil V. > > ____________________ > Racket Users list: > http://lists.racket-lang.org/users ____________________ Racket Users list: http://lists.racket-lang.org/users
