[GHC] #9401: MVar deadlock exceptions cause exceptions in later MVar ops

[GHC]

#9401: MVar deadlock exceptions cause exceptions in later MVar ops
-------------------------------------+-------------------------------------
              Reporter:  snoyberg    |            Owner:  simonmar
                  Type:  bug         |           Status:  closed
              Priority:  normal      |        Milestone:
             Component:  Runtime     |          Version:  7.8.2
  System                             |         Keywords:
            Resolution:  invalid     |     Architecture:  x86_64 (amd64)
      Operating System:  Linux       |       Difficulty:  Unknown
       Type of failure:  Incorrect   |       Blocked By:
  result at runtime                  |  Related Tickets:
             Test Case:              |
              Blocking:              |
Differential Revisions:              |
-------------------------------------+-------------------------------------
Changes (by simonmar):

 * status:  new => closed
 * resolution:   => invalid


Comment:

 This isn't a bug.  There are two threads: the main thread, blocked in
 readMVar, and the child thread, blocked in takeMVar.  BOTH threads are
 deadlocked, so the RTS sends them both an exception.  It has no way to
 tell that unblocking the child thread would later allow the main thread to
 make progress, or indeed whether the reverse might be true.  So the rule
 is simple: when there are a group of deadlocked threads, they are all sent
 an exception simultaneously.

 There's a similar example in my book,
 [http://chimera.labs.oreilly.com/books/1230000000929/ch15.html#sec_deadlock],
 see `deadlock1.hs`.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9401#comment:1>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler