[GHC] #12469: Memory fence on writes to MutVar/Array missing on ARM

[GHC]

#12469: Memory fence on writes to MutVar/Array missing on ARM
-------------------------------------+-------------------------------------
        Reporter:  rrnewton          |                Owner:
            Type:  bug               |               Status:  patch
        Priority:  normal            |            Milestone:  8.0.2
       Component:  Compiler          |              Version:  8.0.1
      Resolution:                    |             Keywords:  memory model
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Incorrect result  |  Unknown/Multiple
  at runtime                         |            Test Case:
      Blocked By:                    |             Blocking:  12537
 Related Tickets:                    |  Differential Rev(s):  Phab:D2495
       Wiki Page:                    |  Phab:D2525
-------------------------------------+-------------------------------------

Comment (by trommler):

 Replying to [comment:27 trommler]:
 > Replying to [comment:25 simonmar]:
 > > If a `sync` fixes it then that would be a useful data point.  If not,
 it means there are barriers missing elsewhere.
 > I had not seen your comment on moving the write barrier.
 >
 > If that was indeed the problem then there is a data dependency and we
 don't have to worry about a load-load barrier on PowerPC (and ARM).
 >
 > I'll do a test build on a PowerPC 970MP and a POWER8.
 On POWER8 I still see a few panics in `mkFastStringWith` and on build
 service (qemu on POWER8 hardware) one panic with `ARR_WORDS object
 entered`. A couple of packages fail with segfaults in `Setup build`.

 The PowerMac (PowerPC 970) build is still running. So far no panic but
 about the same number of packages as before fail in `Setup build`.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12469#comment:28>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler