Significant memory leak

[chkl]

On the server I can currently observer vdiff using 15 GiB (!) of memory. To reproduce:

vdiff --log-level=debug -d /database/2018-04-11-industry-pattern-smart.db --budget=100 --timeout=60 --strategy=smart array_shadowinit_true-unreach-call_true-termination.i 

Some observations:

• It does not occur when using --verifiers="always-Sat"

Current experiment: running with --timeout=1 for faster throughput, and so far I cannot see the memory leak. Maybe the leak is in Timed.hs? (maybe in the custom pipe handling there)...

[chkl]

 valgrind --tool=massif --pages-as-heap=yes vdiff --log-level=debug -d /database/dump.db --budget=100 --timeout=1 --strategy=smart /benchmarks/c/array-industry-pattern/array_shadowinit_true-unreach-call_true-termination.i 

looks like peak memory is ~400K

[chkl]

Using RTS -s, shows for that specific use case indeed very high numbers. Example (Ctrl+C after a few minutes)

6,851,218,724,960 bytes allocated in the heap
  34,397,558,488 bytes copied during GC
   4,078,232,712 bytes maximum residency (23 sample(s))
     271,907,704 bytes maximum slop
            9106 MB total memory in use (485 MB lost due to fragmentation)

                                     Tot time (elapsed)  Avg pause  Max pause
  Gen  0     6465140 colls,     0 par   47.760s  65.971s     0.0000s    3.7486s
  Gen  1        23 colls,     0 par    0.000s   0.002s     0.0001s    0.0017s

  TASKS: 5 (1 bound, 4 peak workers (4 total), using -N1)

  SPARKS: 0 (0 converted, 0 overflowed, 0 dud, 0 GC'd, 0 fizzled)

  INIT    time    0.000s  (  0.001s elapsed)
  MUT     time   59.007s  ( 83.377s elapsed)
  GC      time   47.760s  ( 65.973s elapsed)
  EXIT    time    0.003s  (  0.020s elapsed)
  Total   time  106.770s  (149.370s elapsed)

  Alloc rate    116,109,179,766 bytes per MUT second

  Productivity  55.3% of total user, 55.8% of total elapsed

gc_alloc_block_sync: 0
whitehole_spin: 0
gen[0].sync: 0
gen[1].sync: 0

[chkl]

I think I found the culprit: