I encountered an: internal error: evacuate: strange closure type 803645000

I encountered an: internal error: evacuate: strange closure type 803645000 after loading and running the attached code under GHCi.

Trac metadata

Trac field	Value
Version	7.10.3
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture

mentioned in issue #11108 (closed)

changed weight to 5

added Tbug Trac import labels

Attached file tinychess.hs ($1191).

code that I was running

The last command and output for when this happened. Unfortunately I cannot reproduce.

*Main> :main 6
White to play:
   /---V---V---\
   |   |<K>|   |
   D---+---+---C
   |   |   |   |
   D---+---+---C
   |   | K |   |
   \---A---A---/
[0,3,-5,3,-3,3]
Black to play:
   /---V---V---\
   |   |<K>|   |
   D---+---+---C
   |   |   |   |
   D---+---+---C
   | K |   |   |
   \---A---A---/
[2,5,-3,3,-3,5]
White to play:
   /---V---V---\
   |<K>|   |   |
   D---+---+---C
   |   |   |   |
   D---+---+---C
   | K |   |   |
   \---A---A---/
[0,5,-3,3,-5,3]
Black to play:
   /---V---V---\
   |<K>|   |   |
   D---+---+---C
   |   |   |   |
   D---+---+---C
   |   | K |   |
   \---A---A---/
[-2,3,-3,5,-3,3]
<interactive>: internal error: evacuate: strange closure type 803645000
    (GHC version 7.10.3 for x86_64_unknown_linux)
    Please report this as a GHC bug:  http://www.haskell.org/ghc/reportabug
Aborted

changed milestone to %8.0.1

changed weight to 7

I can also reproduce this with your example. Simply load the file into ghci and run,

$ ghci tinychess.hs 
GHCi, version 7.10.2.20151118: http://www.haskell.org/ghc/  :? for help
[1 of 1] Compiling Main             ( tinychess.hs, interpreted )
Ok, modules loaded: Main.
λ> import Control.Monad (forever)
λ> forever $ playGame 6
.
.
.
   D---+---+---C
   |   | K |   |
   \---A---A---/
[-2,3,-3,5,-3,3]
<interactive>: internal error: evacuate: strange closure type 1195407224
    (GHC version 7.10.2.20151118 for x86_64_unknown_linux)
    Please report this as a GHC bug:  http://www.haskell.org/ghc/reportabug
Aborted
$ 

The problem is evidently still lurking as I see segmentation faults if I run the same test with a recent ghc-8.0 snapshot.

Trac metadata

Trac field	Value
Priority	normal → high

Reproducing in gdb reveals this,

(gdb) bt
#0  0x00007fffecd46507 in __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:55
#1  0x00007fffecd478da in __GI_abort () at abort.c:89
#2  0x00007fffed2f4302 in rtsFatalInternalErrorFn (s=<optimized out>, ap=0x7fffeb38dba8) at rts/RtsMessages.c:182
#3  0x00007fffed2f4460 in barf (s=s@entry=0x7fffed31ab80 "evacuate: strange closure type %d") at rts/RtsMessages.c:46
#4  0x00007fffed2df88e in evacuate (p=p@entry=0x204ed00a8) at rts/dist/build/sm/Evac_thr.c:740
#5  0x00007fffed2fe711 in tidyWeakList (gen=<optimized out>) at rts/sm/MarkWeak.c:270
#6  0x00007fffed2fea08 in traverseWeakPtrList () at rts/sm/MarkWeak.c:130
#7  0x00007fffed300332 in GarbageCollect (collect_gen=collect_gen@entry=1, do_heap_census=do_heap_census@entry=rtsFalse, gc_type=gc_type@entry=2, cap=cap@entry=0x7fffed52ad00 <MainCapability>) at rts/sm/GC.c:409
#8  0x00007fffed2f1d51 in scheduleDoGC (pcap=pcap@entry=0x7fffeb38deb0, task=task@entry=0x7af4f0, force_major=force_major@entry=rtsFalse) at rts/Schedule.c:1652
#9  0x00007fffed2f294e in schedule (initialCapability=<optimized out>, task=task@entry=0x7af4f0) at rts/Schedule.c:551
#10 0x00007fffed2f3a5c in scheduleWorker (cap=<optimized out>, task=0x7af4f0) at rts/Schedule.c:2378
#11 0x00007ffff738c284 in start_thread (arg=0x7fffeb38e700) at pthread_create.c:333
#12 0x00007fffecdfba4d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:109

It appears this is weak-pointer related.

I don't know if this is related or not, but after compiling with profiling enabled:

$ ghc -O2 -fforce-recomp -prof -fprof-auto tinychess.hs

even

$ time ./tinychess 0 +RTS -p

would exhaust memory on my machine and I had to kill it. The argument 0 determines the depth of moves that are considered for evaluating the possible next positions. Without profiling this completes instantly as one would expect.

This appears to be the same issue as #11108 (closed).

Trac metadata

Trac field	Value
Related	- → #11108 (closed)

closed

Lovely. Thanks jme!

Trac metadata

Trac field	Value
Resolution	Unresolved → ResolvedDuplicate

mentioned in commit 9363f04d

added GHCi label

added Phigh label