internal error while running parallel program

Problems occur when I use parFlatMap instead of concatMap.

Appears to be non-deterministic, some kind of subtle threading-related heap corruption.

$ ghc --make -debug -threaded mat_mult_ndp
$ gdb ./mat_mult_ndp
(gdb) run -p 4 test5.mat out.mat +RTS -N4 -DS

mat_mult_ndp: internal error: ASSERTION FAILED: file Sanity.c, line 86

    (GHC version 6.9.20071105 for x86_64_unknown_linux)
    Please report this as a GHC bug:  http://www.haskell.org/ghc/reportabug
[New Thread 1124096336 (LWP 933)]

Program received signal SIGABRT, Aborted.
[Switching to Thread 1107310928 (LWP 931)]
0x0000003e076305b5 in raise () from /lib64/libc.so.6
(gdb) bt
#0  0x0000003e076305b5 in raise () from /lib64/libc.so.6
#1  0x0000003e07632060 in abort () from /lib64/libc.so.6
#2  0x000000000047f888 in rtsFatalInternalErrorFn (s=0x529948 "ASSERTION FAILED: file %s, line %u\n", 
    ap=0x42002d80) at RtsMessages.c:164
#3  0x000000000047f44c in barf (s=0x529948 "ASSERTION FAILED: file %s, line %u\n") at RtsMessages.c:40
#4  0x000000000047f4a6 in _assertFail (filename=0x5390b8 "Sanity.c", linenum=86) at RtsMessages.c:55
#5  0x00000000004b301e in checkClosureShallow (p=0x2aaaaafb0050) at Sanity.c:86
#6  0x00000000004b2efe in checkSmallBitmap (payload=0x2aaaaafb1378, bitmap=2, size=4) at Sanity.c:51
#7  0x00000000004b324c in checkStackFrame (c=0x2aaaaafb1370) at Sanity.c:144
#8  0x00000000004b3418 in checkStackChunk (sp=0x2aaaaafb1360, stack_end=0x2aaaaafb1400)
    at Sanity.c:201
#9  0x00000000004b4a02 in checkTSO (tso=0x2aaaaafb1000) at Sanity.c:715
#10 0x00000000004833fb in threadStackOverflow (cap=0x7872b0, tso=0x2aaaaafb1000) at Schedule.c:2799
#11 0x0000000000481d90 in scheduleHandleStackOverflow (cap=0x7872b0, task=0x7b2bc0, t=0x2aaaaafb1000)
    at Schedule.c:1658
#12 0x00000000004810e2 in schedule (initialCapability=0x7872b0, task=0x7b2bc0) at Schedule.c:694
#13 0x0000000000482f3b in workerStart (task=0x7b2bc0) at Schedule.c:2528
#14 0x0000003e096061c5 in start_thread () from /lib64/libpthread.so.0
#15 0x0000003e076d062d in clone () from /lib64/libc.so.6

I tested it on an older install of GHC 6.7.20070831 and it had the same problem. test5.mat is attached sample input describing two matrices of size nxn (=64 in this case). Smaller inputs didn't seem to tickle the bug, or at least, not often enough to be noticed.

Trac metadata

Trac field	Value
Version	6.9
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Runtime System
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture