[GHC] #7076: Crash with eager blackholing

[GHC]

#7076: Crash with eager blackholing
------------------------------+---------------------------------------------
 Reporter:  SimonMeier        |          Owner:                   
     Type:  bug               |         Status:  new              
 Priority:  normal            |      Component:  Runtime System   
  Version:  7.4.2             |       Keywords:  eager blackholing
       Os:  Unknown/Multiple  |   Architecture:  x86_64 (amd64)   
  Failure:  Runtime crash     |       Testcase:                   
Blockedby:                    |       Blocking:                   
  Related:  #6146             |  
------------------------------+---------------------------------------------
 In the context of the Tamarin prover (https://github.com/tamarin-prover
 /tamarin-prover), I am consistently getting segfaults with -feager-
 blackholing and -fthreaded both switched on. Without eager blackholing
 (i.e., just -fthreaded), these segfaults are gone.

 The crash only occurs only when using the GUI, which is implemented as an
 integrated Yesod web application. Thus the features that we use are
 Template Haskell, Quasi Quoting, and calling to system processes like the
 http://www.graphviz.org/ 'dot' tool for rendering graphs. We also use one
 'unsafePerformIO' for interacting with an instance of another process that
 we use as an AC-unification oracle. This interaction is protected by an
 'MVar' and has been working for over a year now. I'm just mentioning it to
 simplify thinking about the possible reasons why -feager-blackholing leads
 to segfaults.

 The codebase of the prover is rather large and I'm sorry that I cannot
 provide a small test case. Are there any options to find out more
 precisely where the program crashes?

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/7076>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler