From: Beman Dawes (bdawes_at_[hidden])
Date: 2001-01-29 13:47:16
At 01:26 PM 1/29/2001 -0500, Jeremy Siek wrote:
>I have a concern about the requirement that the tests must execute
>For example, the current graph test only tests one version of the
>adjacency_list class, but I was planning on adding code this week to make
>sure all the versions of adjacency_list get tested in the regression
>The problem is that there are 48 different versions of adjacency_list.
>Also, when I finally get unit tests in place for all of the algorithms,
>there will be another 40 or so tests.
>Should the regressions test be divided into a "quick test" and
>a "full test"?
The "quick" or "basic" tests need to establish a level of confidence, but
just can't go on very long.
On some libs, a "full test" may be plenty quick enough for inclusion in the
main regression test. So maybe we can find a better name for those.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk