[GHC] #10368: STM test failing on Armhf/Linux (was: Armhf/Linux : test T7815 failing)

[GHC]

#10368: STM test failing on Armhf/Linux
-------------------------------------+-------------------------------------
        Reporter:  erikd             |                   Owner:
            Type:  bug               |                  Status:  new
        Priority:  normal            |               Milestone:  7.12.1
       Component:  Compiler          |                 Version:  7.11
      Resolution:                    |                Keywords:
Operating System:  Unknown/Multiple  |            Architecture:  arm
 Type of failure:  Incorrect result  |               Test Case:
  at runtime                         |                Blocking:
      Blocked By:                    |  Differential Revisions:
 Related Tickets:  #7815             |
-------------------------------------+-------------------------------------
Description changed by erikd:

Old description:

> During validation on armhf/linux, I found this test had failed.
>

> Unfortunately, it only fails intermittently on one quad core Arm board
> and not at all on another quad core Arm board. If I do 10 runs of the
> test like:
>
> {{{
> for x in $(seq 1 10) ; do testsuite/tests/rts/T7815 50000 +RTS -N2 ; echo
> $? ; done
> }}}
>
> one will fail at least 4 or 5 times and ocassionally as many as 9 or 10
> times.
>
> The two boards are:
>
> * Inforce Computing ifc6540 with a Qualcomm Snapdragon 805 CPU.
> * Radxa Rock with a Rockchip RK3199 CPU.
>
> The ifc6540 is the one that fails.

New description:

 During validation on armhf/linux, I found that test T7815 had failed.


 Unfortunately, it only fails intermittently on one quad core Arm board and
 not at all on another quad core Arm board. If I do 10 runs of the test
 like:

 {{{
 for x in $(seq 1 10) ; do testsuite/tests/rts/T7815 50000 +RTS -N2 ; echo
 $? ; done
 }}}

 one will fail at least 4 or 5 times and ocassionally as many as 9 or 10
 times.

 The two boards are:

 * Inforce Computing ifc6540 with a Qualcomm Snapdragon 805 CPU.
 * Radxa Rock with a Rockchip RK3199 CPU.

 The ifc6540 is the one that fails.

 @fryguybob suggests that this is actually a bug in the STM implementation
 that breaks on Arm because of Arm's weaker memory consistency model.

--

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10368#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler