I think Schwern's proposal of a patch for every test, applied appropriately
will do a good job of making sure the Perl test suite has close to 100%
functional coverage of the core. However, we should also consider doing
code coverage analysis. This can help us to create test cases that
accurately test the fringe cases.
We could actually do a run on perl5 with the JART list and/or CPAN list to
see how thoroughly those proposed regression suites would test it. If it
scores less than 50% of the code, we know we need yet another suite of
(probably more focused) regression tests.
++t
