We only really need to have the canonical path for comparison purposes in one place, and I've heard that it caused a bug for someone when it was always canonicalized.
This should eliminate an infinite loop bug reported by travitch, which occurs when a symlink causes the two paths to appear different even when they point to the same file.
Hopefully the inclusion of the FFI on other platforms doesn't cause any major issues.
It's mainly a confidence thing for me. I don't trust my cobbled- together tree of scripts enough to really believe it when it says "All tests passed". So now, all input to the assert() function is displayed on stderr as well. Now I trust the results of my tests.
* Use HLint to tweak code. * Make the test suite more comprehensive
This is a fairly significant internal logic change, so it feels dishonest to just mark it as a minor revision.
Replaced 'getReconf' with 'getForceReconf'
This is basically so FreeBSD will semi-work until Balazs can get some new BSD support code into the 'executable-path' library.
This parameter allows an application to define special entry points which will be used to pass control directly through to their main function, rather than doing all the usual recompile checks. In short, it allows applications to bypass all of Dyre's recompilation functions, only passing through a small amount of housekeeping before entering the main application.