- move computation of nullspace to splu, correct problem of choice of
algorithm
- create config files with repository URLs etc, don't
hard-code URLs in scripts
- test and update all create_*_release scripts for all sub-projects
- test build_package
- replace loop over svn-commands on failure with a simple logging-and-retry
functionality (the $status test in while fails, because end exits with
status 0 before while is executed again -> only one repetition!) See
update_admin for basic solution.
- add support for (linear) matrix problems (for example, to reduce system
size for variational problems)
%COCO Execute a continuation run.
%
% BD = COCO([PROB,] RUN, TB_SEL, CONT_ARGS)
% Execute a continuation run. The resulting data will be saved under run
% name RUN. The zero problem is defined by the toolbox selected with the
% arguments passed as TB_SEL. The arguments provided as CONT_ARGS are
% passed on to the constructor of the covering toolbox.
%
% TB_SEL = { TB_NAME, ISOL_T, SOL_T, TB_ARGS }
% Use zero problem from toolbox TB_NAME. The constructor for the zero
% problem is the function with name sprintf('%s_%s2%s', TB_NAME, ISOL_T,
% SOL_T) and constructs a zero problem starting at a solution of type
% ISOL_T and computing a branch of solutions of type SOL_T. Arguments
% required by the problem constructor are passed as TB_ARGS.
%
% TB_SEL = { @TB_CTOR, TB_ARGS }
% Use zero problem constructed by the problem constructor TB_CTOR. Arguments
% required by the problem constructor are passed as TB_ARGS.
%
% TB_SEL = { [] }
% Use zero problem defined in PROB. With this syntax, the argument PROB is
% compulsory.
%
% See also: COVERING_CREATE, COCO_BD_READ, COCO_BD_PRINT, COCO_PROB,
% COCO_SET
% Copyright (C) Frank Schilder, Harry Dankowicz
% $Id$