Runtime caught in an infinite loop

[lpsmith]

Triggering this bug seems a little fragile; but the program and method I've attached triggers the bug every time for me. It is minimal-ish, I certainly have cut it down a lot, and it perhaps could be made even smaller, but a lot of seemingly irrelevant changes to the file will fail to trigger the bug.

$ cabal configure && cabal build
$ ./dist/build/deadlock/deadlock

And then edit the file and change the 1 to some other value, like 2. Configurator will reload the file within a second, and then I see that the thread that periodically writes something to standard out stops working, and that CPU usage goes to 100%, and that I can no longer type something in and have the program echo it back to me.

GHC-7.0.3 has no problem running this program.

[lpsmith]

Ok, I think I have a better fix on what is going on now; I finally managed to replicate the bug in a program I specifically wrote to trigger this bug, instead of one I cut down from the actual program I am writing.

If I get some time tomorrow, I'll see if I can't study the code path triggered inside of configurator and develop an example that doesn't depend directly on it. But I think I can rule out a library bug, as it works fine on GHC 7.0.3 with the same libraries and an infinite nonproductive loop outside of an unsafe FFI call shouldn't block the other Haskell threads.

[lpsmith]

This bug may be a confluence of #5784 and #5783.

[lpsmith]

Yes, I do believe it is a combination of the bugs above.