Known Bugs: Difference between revisions

From XDSwiki
Jump to navigation Jump to search
No edit summary
No edit summary
Line 1: Line 1:
* The October 2007 and the November 2007 versions do not evaluate DATA_RANGE= in the CORRECT step. The August 2007, and versions since December 2007 do.
* The October 2007 and the November 2007 versions do not evaluate DATA_RANGE= in the CORRECT step. The August 2007 version, and versions since December 2007 do.
* The June and July 2008 versions of "xds_par" under Linux suffer from a compiler bug in the CORRECT step: the correlation coefficient against a reference data set is always 0.00, and therefore sometimes the wrong decision is made concerning re-indexing of the data. The workaround is to use the single-processor version ("xds") instead.
* The June and July 2008 versions of "xds_par" under Linux suffer from a compiler bug in the CORRECT step: the correlation coefficient against a reference data set is always 0.00, and therefore sometimes the wrong decision is made concerning re-indexing of the data. Example taken from CORRECT.LP produced with xds_par:
CORRELATION  NPAIR  Rmeas  COMPARED  ESD      REINDEX TRANSFORMATION
      0.00211818312  25.1  53803    1.65  1  0  0  0  0  1  0  0  0  0  1  0
  *  0.00260052808  25.1  53803    1.65  1  1  0  0 -1  0  0  0  0  0  1  0
whereas xds produces the correct output:
CORRELATION  NPAIR  Rmeas  COMPARED  ESD      REINDEX TRANSFORMATION
  *  0.89    5960  25.1  53803    1.65  1  0  0  0  0  1  0  0  0  0  1  0
      0.06    5960  25.1  53803    1.65  1  1  0  0 -1  0  0  0  0  0  1
The workaround is to use the single-processor version ("xds") instead.

Revision as of 10:14, 5 September 2008

  • The October 2007 and the November 2007 versions do not evaluate DATA_RANGE= in the CORRECT step. The August 2007 version, and versions since December 2007 do.
  • The June and July 2008 versions of "xds_par" under Linux suffer from a compiler bug in the CORRECT step: the correlation coefficient against a reference data set is always 0.00, and therefore sometimes the wrong decision is made concerning re-indexing of the data. Example taken from CORRECT.LP produced with xds_par:
CORRELATION  NPAIR  Rmeas  COMPARED  ESD       REINDEX TRANSFORMATION

     0.00211818312   25.1   53803    1.65  1  0  0  0  0  1  0  0  0  0  1  0
 *   0.00260052808   25.1   53803    1.65  1  1  0  0 -1  0  0  0  0  0  1  0

whereas xds produces the correct output:

CORRELATION  NPAIR  Rmeas  COMPARED  ESD       REINDEX TRANSFORMATION

 *   0.89     5960   25.1   53803    1.65  1  0  0  0  0  1  0  0  0  0  1  0
     0.06     5960   25.1   53803    1.65  1  1  0  0 -1  0  0  0  0  0  1 

The workaround is to use the single-processor version ("xds") instead.