I haven't checked your program, but what you have textually described is
#8680: In STM: Variables only in left branch of orElse can invalidate the right branch transaction --------------------------------------------+------------------------------ Reporter: jberryman | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.6.3 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Runtime performance bug | Unknown/Multiple Test Case: | Difficulty: Unknown Blocking: | Blocked By: | Related Tickets: --------------------------------------------+------------------------------ Comment (by jberryman): Replying to [comment:2 ezyang]: the behavior I would expect for orElse (and is consistent with the semantics): the 'else' branch is only valid if the 'if' branch retries; so if the first branch is invalidated, it may not be retrying anymore. So I think I got caught up in the details and convinced myself that there wasn't really a way to do any useful reasoning with this behavior. But I suppose it lets you say e.g. "if we're in the right branch then our view of the world is one where predicate P from the left branch is False", even if the two branches share no variables and the first is rolled back.
This suggests there might be a useful nondeterministic version of orElse, which relaxes this restriction.
Now I'm not even sure that the current `orElse` behavior won't work for me. Would it be better to update this ticket when I have a clearer idea, or just consider opening a new feature request ticket in the future? Thanks for the feedback. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/8680#comment:3 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler