[GHC] #13148: Adding weak pointers to non-mutable unboxed values segfaults

[GHC]

#13148: Adding weak pointers to non-mutable unboxed values segfaults
-------------------------------------+-------------------------------------
        Reporter:  mboes             |                Owner:
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.0.1
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by rwbarton):

 I don't know exactly how `mkWeakNoFinalizer#` works but surely it can only
 possibly attach a finalizer to a value of ''boxed'' type, that is, a value
 that is stored as a closure on the Haskell heap. After all the meaning of
 a finalizer is that it's an action that runs when the object is no longer
 live after GC. `MutVar#` and `MVar#` are boxed (but unlifted) types, but
 `Addr#` is unboxed, really just a synonym for `Int#`. It doesn't make any
 more sense to attach a finalizer to `p#` than it does to attach one to
 `7#`. See
 https://ghc.haskell.org/trac/ghc/wiki/Commentary/Rts/Storage/HeapObjects
 for the difference between unboxed and unlifted.

 The segfault happened because the weak finalizer code interpreted the bit
 pattern of `p#` as a pointer into the Haskell heap, while it was actually
 a pointer into the C heap (produced by `malloc`); and chaos ensued. (Based
 on this program snippet, it looks like you might just want
 `mallocForeignPtr`.) But morally the problem isn't really with C vs.
 Haskell heap but with trying to attach a finalizer to an unboxed value.

 The documentation you refer to is the documentation for the user-facing
 System.Mem.Weak module. We could add documentation to
 `mkWeakNoFinalizer#`, even though the primops are really only intended to
 be used to implement the base libraries. What would have been helpful?

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13148#comment:1>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler