MrBayes2.01 15 August 2001 (c) John Huelsenbeck & Fredrik Ronquist 1. Known problems (version 2.01) The program will not interpret information after the first MrBayes block correctly. This will be corrected in version 3. For now, make sure there is only one MrBayes block in each batch file. The 'sumt' and 'sump' commands are not able to deal with '.t' and '.p' files that contain information from several runs. Thus, if you append the results from one run to those of other runs, you need to delete all the information from the previous run(s) before running 'sumt' or 'sump' on the '.t' or '.p' file. In version 3, 'sumt' and 'sump' will summarize the information in the last block in the '.t' or '.p' file. The program does not handle execute statements in batch files correctly. The file that is called is executed correctly but when returning to the calling file the command parser inappropriately thinks it is outside of any block. This problem can be solved by inserting 'begin mrbayes;' after the execute statement in the calling file. This will be corrected in version 3. After the program executes a 'sumt' or a 'sump' command in a batch file, the following lines are interpreted incorrectly. Again, this problem can be solved by inserting 'begin mrbayes;' after the 'sumt' or 'sump' command. This will be corrected in version 3. 2. Version history Version 2.01 ---------------------------------- The #NEXUS file header is now printed correctly to output files. A problem with the invariant and invariant gamma models in all previous versions of the program is now corrected. The bug caused long chains to stop with the error message "WARNING: In LIKE_EPSILON". Memory leaks introduced in version 2.0 have been deleted. The usual symptom of these bugs would be termination of a chain after many generations with an error message such as "ERROR: Problem allocating " and then the name of some variable. The tokens in the translate command are now printed in the correct order in the file containing the consensus tree (the .con file). Several bugs in the likelihood functions for the standard and restriction models are now fixed. The bugs would crash the program when these models were invoked. A bug that caused the program to crash when it was asked to open a non-existing file is now fixed. Version 2.0 ----------------------------------------- All likelihood functions are completely rewritten, resulting in significant speed gains for most data types. Models for restriction sites and for standard binary characters have been added. Parsimony model (Tuffley & Steel) has been added. New topology and branch length move types have been introduced. User can set proposal probabilities of different move types. The sumt command outputs a consensus tree with posterior clade probabilities in addition to partition frequencies. If branch lengths are saved, a consensus phylogram is output as well. 3. Bug reports Send bug reports to John Huelsenbeck or Fredrik Ronquist . =============================== Plan 9 port by Peter A. Cejchan Mon Dec 9 06:53:19 EST 2002 Changes in sources (if any) marked '++pac'. Type 'mk install' to compile and install. Enjoy!