home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.software-eng
- Path: sparky!uunet!spool.mu.edu!wupost!cs.uiuc.edu!marick
- From: marick@cs.uiuc.edu (Brian Marick)
- Subject: Testing bug fixes - request for data
- Message-ID: <C099nG.BD9@cs.uiuc.edu>
- Organization: University of Illinois, Dept. of Comp. Sci., Urbana, IL
- Date: Sun, 3 Jan 1993 02:02:03 GMT
- Lines: 59
-
-
- I'd like to start a large-scale, low-key, industry-specific research
- project in testing bug fixes.
-
- I have a technique for testing subsystems (small chunks of code up to,
- say, 10,000 lines) which is sketchily described in parts of several
- papers I'll reference below and completely described in a book I'm
- writing. It's mostly variations on, and additions to, standard
- testing techniques. By applying the technique to real programs
- (especially my own), I think I've refined it to a practical balance of
- cost and effectiveness.
-
- What I am shakier on is the efficient testing of bug fixes, particularly
- in the absence of a good regression test suite. I'm interested in two
- types of testing:
-
- - checking that a change actually fixes the bug.
- - checking whether the bug fix induces new bugs because of
- unanticipated dependencies.
-
- (The latter is more interesting, because harder.)
-
- For effective testing, I could test the entire containing subsystem
- using the complete technique, but that's not a realistic option: too
- expensive. What I need to do is specialize the technique for
- bugfixes, make it good at finding likely problems but not too
- expensive to apply to a small change.
-
- I have approaches, but they must be tuned with real data, actual
- bugfixes that were incorrect or caused new bugs. ("Tuning", of
- course, includes "throwing out and replacing".) Since I work for
- myself, I don't have access to big bug databases. So I thought I'd
- ask those who do to give me a peek into theirs. I'll of course sign
- non-disclosure agreements.
-
- You benefit by getting my advice on how that bug could have been
- found, by eventual publication of articles or a book that will help
- your testing, and by getting credit in anything I publish. The
- community wisdom about testing bug fixes will grow faster, to
- everyone's benefit.
-
- What do you have to do? When you fix a bug, and that bug turns out to
- have been caused by a previous bug fix, you need to describe the first
- (bug-causing) change and what was wrong about it. I've done this to a
- limited extent with one client already. These types of bugs take time
- to explain (diffs often don't suffice), so you must be willing to
- spend up to an hour per bug (usually less). Explaining in person can
- be more efficient. I'll be in the San Jose area a lot in 1993, and I
- get out to Boston every once in a while.
-
- If anyone has any helpful literature citations, I'd like those as well.
-
- I expect to spend a few years on this.
-
- Thanks.
-
- Brian Marick, marick@cs.uiuc.edu, testing!marick@uunet.uu.net
- Testing Foundations: Consulting, Training, Tools.
- Freeware test coverage tool: see cs.uiuc.edu:pub/testing/GCT.README
-