martijn.fg
    junction tree builds too slow (not really a bug, but maybe an 
    undesirable feature)

bugtrw.fg
    tests/errorbounds/logz bugtrw.fg TRW 0 3
    will fail because of NANs:
        iter = 52, cur_logZ = 35.08234633, weight = 3.615346277
        checking alpha = 0.01584706616...
        Starting TRW[maxiter=1000,maxouteriter=100,outertol=1e-06,tol=1e-09,updates=SEQFIX,verbose=3]:inner_loop ...converged in 70 passes (25 clocks).
          for alpha = 0.01584706616, gain = 0.0002234145247 >=? alpha * weight = 0.05729263163, gain / (alpha * weight) = 0.003899533296
              *0.01584706616 0.0002234145247
        checking alpha = 0.007923533079...
        Starting TRW[maxiter=1000,maxouteriter=100,outertol=1e-06,tol=1e-09,updates=SEQFIX,verbose=3]:inner_loop ...converged in 69 passes (24 clocks).
          for alpha = 0.007923533079, gain = 0.0001428102949 >=? alpha * weight = 0.02864631581, gain / (alpha * weight) = 0.004985293602
               0.007923533079 0.0001428102949
              *0.01584706616 0.0002234145247
        checking alpha = 0.5079235331...
        Starting TRW[maxiter=1000,maxouteriter=100,outertol=1e-06,tol=1e-09,updates=SEQFIX,verbose=3]:inner_loop ...converged in 114 passes (40 clocks).
          for alpha = 0.5079235331, gain = nan >=? alpha * weight = 1.836319454, gain / (alpha * weight) = nan
               0.007923533079 0.0001428102949
              *0.01584706616 0.0002234145247
               0.5079235331 nan
        logz: trw.cpp:341: virtual double libDAI::TRW::run(): Assertion `0.0 < alpha' failed.


ldpcrsl.fg
    LCBP_UNICAV_SEQRND and LCBPH_UNICAV_SEQRND yield irreproducible results:

        cn02> ./test_ldpc --filename /tmp/ldpcrsl.fg --aliases ldpc.aliases --methods JTREE_HUGIN LDPC[] BP_SEQMAX GBP_MIN LCLIN_UNICAV_SEQRND LCLIN_UNICAV_SEQRND LCBP_UNICAV_SEQRND LCBP_UNICAV_SEQRND LCBPH_UNICAV_SEQRND LCBPH_UNICAV_SEQRND
        # /tmp/ldpcrsl.fg
        # METHOD                                      CLOCKS     MAX ERROR   AVG ERROR  LOGZ ERROR     MAXDIFF
        JTREE_HUGIN                                        1    
        LDPC[]                                             0     9.574e-04   7.582e-05  -1.000e+00   1.000e-09
        BP_SEQMAX                                          0     5.922e-05   1.211e-05   1.182e-05   1.000e-09
        GBP_MIN                                            1     5.922e-05   1.211e-05   1.182e-05   1.000e-09
        LCLIN_UNICAV_SEQRND                               19     5.922e-05   1.211e-05         nan   1.000e-09
        LCLIN_UNICAV_SEQRND                               17     5.922e-05   1.211e-05         nan   1.000e-09
        LCBP_UNICAV_SEQRND                                30     6.509e-01   2.931e-01         nan   1.000e-09
        LCBP_UNICAV_SEQRND                                31     7.484e-01   3.852e-01         nan   1.000e-09
        LCBPH_UNICAV_SEQRND                             1585     2.262e-04   5.626e-05         nan   1.000e-09
        LCBPH_UNICAV_SEQRND                             3288     5.751e-05   1.160e-05         nan   3.247e-09


ldpc.fg
    ALL LC*_UNICAV_SEQRND yield irreproducible results:
    
        cn02> ./test_ldpc --filename /tmp/ldpc.fg --aliases ldpc.aliases --methods JTREE_HUGIN LDPC[] BP_SEQMAX GBP_MIN LCLIN_UNICAV_SEQRND LCLIN_UNICAV_SEQRND LCBP_UNICAV_SEQRND LCBP_UNICAV_SEQRND LCBPH_UNICAV_SEQRND LCBPH_UNICAV_SEQRND
        FactorGraph::FactorGraph():  WARNING: short loops are present
        # /tmp/ldpc.fg
        # METHOD                                      CLOCKS     MAX ERROR   AVG ERROR  LOGZ ERROR     MAXDIFF
        JTREE_HUGIN                                        1    
        LDPC[]                                             0     9.574e-04   7.582e-05  -1.000e+00   1.000e-09
        BP_SEQMAX                                          0     9.574e-04   7.582e-05   1.304e-04   1.000e-09
        GBP_MIN                                            0     5.922e-05   1.211e-05   1.182e-05   1.000e-09
        LCLIN_UNICAV_SEQRND                               11     1.577e-03   1.171e-04         nan   1.000e-09
        LCLIN_UNICAV_SEQRND                               13     3.792e-03   2.648e-04         nan   1.000e-09
        LCBP_UNICAV_SEQRND                                22     5.813e-01   3.001e-01         nan   1.000e-09
        LCBP_UNICAV_SEQRND                                34     9.165e-01   4.073e-01         nan   1.000e-09
        LCBPH_UNICAV_SEQRND                             2851     1.777e-02   3.419e-03         nan   1.160e-03
        LCBPH_UNICAV_SEQRND                             1986     5.919e-05   1.170e-05         nan   1.000e-09

test_dreg_254089601.fg
    LCBP2
        gives nan

test_dreg_-1448249693.fg
    LCBP2
        test: prob.h:375: T TProb<T>::normalize(TProb<T>::NormType) [with T = Real]: Assertion `Z != 0.0' failed.

test_dreg_1878472930.fg
    LCBP2
        test: prob.h:375: T TProb<T>::normalize(TProb<T>::NormType) [with T = Real]: Assertion `Z != 0.0' failed.

test_dreg_2235838938.fg
    TREEEP
        test: prob.h:375: T TProb<T>::normalize(TProb<T>::NormType) [with T = Real]: Assertion `Z != 0.0' failed.

test_dreg_3732542421.fg
    TREEEP
        test: prob.h:375: T TProb<T>::normalize(TProb<T>::NormType) [with T = Real]: Assertion `Z != 0.0' failed.

potent129.fg, potent129_rsl.fg
    LCBP*_PAIRCAV_*:
        LCBP::run:  initial _pancakes has NaNs!

test_160976619.fg
    LCBPK_PAIRCAV_SEQFIX
        test: prob.h:429: TProb<T> TProb<T>::normalized(TProb<T>::NormType) const [with T = double]: Assertion `Z != 0.0' failed.

bugcvm.fg
    HAK
        gives uniform distribution, but it shouldn't
