[GHC] #8006: Asynchronous exception rethrown synchronously inside runStmt

[GHC]

#8006: Asynchronous exception rethrown synchronously inside runStmt
---------------------------------+------------------------------------------
    Reporter:  edsko             |       Owner:                  
        Type:  bug               |      Status:  new             
    Priority:  normal            |   Milestone:                  
   Component:  Compiler          |     Version:  7.6.3           
    Keywords:                    |          Os:  Unknown/Multiple
Architecture:  Unknown/Multiple  |     Failure:  None/Unknown    
  Difficulty:  Unknown           |    Testcase:                  
   Blockedby:                    |    Blocking:                  
     Related:                    |  
---------------------------------+------------------------------------------

Comment(by simonmar):

 If repeating the I/O is what you want, then you can do that - indeed,
 that's what I suggested that we should do, and this ticket is Edsko's
 followup after investigating that possibility.  For some background, we
 adopted this approach in the I/O library for lazy I/O, see #3997, and
 `Note [async]` in `libraries/base/GHC/IO/Handle/Internals.hs`.

 Edsko's objection to doing this in GHC is that there are a lot of places
 where we catch exceptions inside `unsafePerformIO` or equivalent, and it
 probably isn't safe to repeat the I/O in all of them.  Well, maybe it is -
 I'm not sure.

 The semantics of async exceptions doesn't say anything about this, because
 it doesn't include `unsafePerformIO` - IO is strictly at a higher level
 than pure computation.

 What seems attractive about having a different kind of catch is that it
 avoids the problem in cases where we're only interested in catching a
 subset of exceptions, which is what we're doing in GHC: both `tryM` and
 `tryMostM` want to just pass through async exceptions.  You're right that
 it doesn't help if you want to do some cleanup (`bracket` or `finally`).

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/8006#comment:8>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler