opencv-2.2 with TBB uses only one CPU-core/opencv-2.0 with openmp uses all four, but seg fault (Bug #1068)

Hello,

I've downloaded opencv-2.2 release from willowgarage.com, compiled
this way: mkdir release && cd release && cmake D CMAKE_BUILD_TYPE=RELEASE -DWITH_TBB=ON ..
and even got
"- Use TBB: YES" in the output.

However when I run time /usr/local/bin/opencv_haartraining -data trainout -vec positives.vec -bg negatives.txt -npos 110 -nneg 1000, it uses only one CPU-core. Operating system - ubuntu-10. btw, opencv_haartraining from the default opencv-dev package on ubuntu-10 also uses only one core.

I also tried the version from trunk, no luck - it still uses only one CPU.
However when I compiled opencv-2.0 (with openmp), it uses all 4 cores - however I got a segmentation fault:

time /usr/local/bin/opencv-haartraining -data trainout -vec positives.vec -bg negatives.txt -npos 110 -nneg 1000
Data dir name: trainout
Vec file name: positives.vec
BG file name: negatives.txt, is a vecfile: no
Num pos: 110
Num neg: 1000
Num stages: 14
Num splits: 1 (stump as weak classifier)
Mem: 200 MB
Symmetric: TRUE
Min hit rate: 0.995000
Max false alarm rate: 0.500000
Weight trimming: 0.950000
Equal weights: FALSE
Mode: BASIC
Width: 24
Height: 24
Applied boosting algorithm: GAB
Error (valid only for Discrete and Real AdaBoost): misclass
Max number of splits in tree cascade: 0
Min number of positive samples per cluster: 500
Required leaf false alarm rate: 6.10352e-05
Stage 0 loaded
Stage 1 loaded
Stage 2 loaded
Stage 3 loaded
Stage 4 loaded
Stage 5 loaded
Stage 6 loaded

Tree Classifier
Stage
----------+---+---+---+ | 0| 1| 2| 3| 4| 5| 6|
----------+---+---+---+

0---1---2---3---4---5---6

Number of features used : 85848

Parent node: 6

• 1 cluster ***
  POS: 46 46 1.000000
  NEG: 418 0.0210856
  BACKGROUND PROCESSING TIME: 0.00
  Precalculation time: 0.00
  ------+-+---------+---------+---------+---------+ | N |%SMP|F| ST.THR | HR | FA | EXP. ERR|
  ------+-+---------+---------+---------+---------+ | 1|100%|-|-0.395958| 1.000000| 1.000000| 0.254310|
  ------+-+---------+---------+---------+---------+ | 2|100%|+|-1.186469| 1.000000| 1.000000| 0.174569|
  ------+-+---------+---------+---------+---------+ | 3|100%|-|-0.865389| 1.000000| 0.868421| 0.351293|
  ------+-+---------+---------+---------+---------+ | 4| 88%|+|-1.444954| 1.000000| 0.868421| 0.318966|
  ------+-+---------+---------+---------+---------+ | 5| 77%|-|-0.444954| 1.000000| 0.746411| 0.594828|
  ------+-+---------+---------+---------+---------+ | 6| 77%|+|-1.101890| 1.000000| 0.751196| 0.278017|
  ------+-+---------+---------+---------+---------+
  Segmentation fault

real 0m9.415s
user 0m27.270s
sys 0m0.068s

It would so great to run haartraining on many CPUs, for example on cx1g.large amazon ec2 with 16 cores.

Thanks.