#4 Best between NNI and SPR is still broken

[Marcelo S Zanetti]

codonphyml-omp -i cymon.phy -d codon -m MG -g THBAPPC -t e -s BEST -f optimize -w g --wclasses 6

. (179696 sec) [ -300178.5019] [ts/tv ratio ][3.14 ]
. (179696 sec) [ -300178.5019] [dn/ds ratio ]
. (179696 sec) [ -300178.5019] [Alpha and Beta ][0.68 1.86 ]
. (179696 sec) [ -300178.5019] [CF3x4 freqs. ]
. (179741 sec) [ -300178.5019] [Topology ]
. (180547 sec) [ -300178.5014] [Topology ]
. (181803 sec) [ -300178.5014] [ts/tv ratio ][3.14 ]
. (181803 sec) [ -300178.5014] [dn/ds ratio ]
. (181803 sec) [ -300178.5014] [Alpha and Beta ][0.68 1.86 ]
. (181803 sec) [ -300178.5014] [CF3x4 freqs. ]

. Checking for NNIs, optimizing five branches...

. best_lnL = -300178.496325, c_lnL = -300178.501354
. Err in file utilities.c at line 11530

Testing with -s NNI and -s SPR works fine ... the problem is the -s BEST

[Marcelo S Zanetti]

Commented out two sanity checks in file utilities.c, within function "Best_Of_NNI_And_SPR". It is similar to the numerical issues we had during execution, while restoring NNI and SPR moves. The error comes from that fact that the original likelihood is slightly different from the number obtained after restoring a move. Must be a numerical issue due to large tree sizes, with long sequences.