This is the Release Candidate for Version 3.9
It incorporates improvements to the logic of COBCD-ST, including more robust handling of REDEFINES and some 66/RENAMES capability. COBCD-ST is significantly faster because of the elimination of a couple of O(N-squared) constructions in the code. For a program with 40,000 COBOL variables it is thirty times faster. COBCD-SFIX is faster by a factor of about six. The Python cobcd.py has been altered to pick up the cross-reference VARIABLE_STRING much faster than before. A test program with 40,000 COBOL variables used to take 2.5 minutes to read over the string, one character at a time. Use of a different GDB command for looking at string data got that down to 1.5 seconds.