It seems to me that the precompiler doesn’t lock files properly to prevent race conditions on generating precompilation files. I have a script which uses the Distributed module which I run on a single multicore computer running Linux, so all the worker threads share the same precompilation cache. The script executes many lines of the form
@everywhere using Module_AAA
and I have the distinct impression that the precompiler of v. 1.1 doesn’t like this (lots of precompilation errors are generated). If I precompile all included modules by executing a non-parallel script consisting of only the same “using” statements without @everywhere, the precompiler works ok, and then the parallel script subsequently runs without precompilation errors (since all modules are up-to-date in the precompilation cache).