> what is our loss in confidence in the correctness of an algebra result > as a consequence of changes in the current codebase? Do we have > confidence to begin with?
re: Algebra change confidence After initial merge failures creating silver, a new facility was created to check what changes occur in any file when a change is made (see src/regress/REGRESS). Then, starting with Gold, each change was applied, one at a time, and a complete system build per change was done. The REGRESS script computes an MD5 hashcode for each file and remembers the results. After each change is checked, the new results file replaces the prior one. The MD5 hash will be the same if no change occurs. We can see whether a change has had any effect on the algebra. Indeed we can now see every file that a patch affects. All patches were applied that did not affect the algebra. Changes that affected the algebra were not yet applied. This gives us great confidence that nothing in the algebra is broken (any worse than it was in Gold). re: Algebra correctness. To look at actual computation results there is a new regress function that does a line-by-line compare against prior results cached in the comments. This has uncovered interesting problems going from platform to platform which we need to think about. More tests are being added as well as documentation about each test. The src/input subdir will eventually contain CATS-like test cases for the range of algebra along with associated documentation. Prior to this the input files were simple test scripts. After this the expected results are explained mathematically and then the results from Axiom are given and tested. Tim _______________________________________________ Axiom-developer mailing list [email protected] http://lists.nongnu.org/mailman/listinfo/axiom-developer
