On Mon, 2005-08-22 at 20:15 +0200, Johannes Schmidt-Fischer wrote: > Dave Shield wrote:
> > In general, we've tended to regard each test as completely independent > > of the others. In principle, it'd be perfectly possible to run *just* > > tests 48 and 49 (without running 47 first) So I'd be reluctant to > > have the results of one test rely on a previous one. > > Do you mean "not rely on a previous one"? That way I can use every test > independent of any other (and in any order if necessary). That's right. As far as the *design* of the overall test suite is concerned, it's probably reasonable for later tests to assume that earlier tests succeeded. For example, if one of the first tests exercises SNMPv2c GET requests with certain core objects, then a later test (perhaps to test the handling of Counter64 objects) can assume that SNMPv2c GET requests work, so any failure will indicate a problem with 64-bit processing. A full test run with a broken "snmpget" command would result in *all* the tests failing - but it would be up to the person running the tests to ignore the later reports, and recognise the early problems as the fundamental cause. But the test should be standalone, in that it should be possible to pick and choose among them (and in any order). Dave ------------------------------------------------------- SF.Net email is Sponsored by the Better Software Conference & EXPO September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf _______________________________________________ Net-snmp-coders mailing list Net-snmp-coders@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/net-snmp-coders
