Op 3 aug. 2013 21:56 schreef "Scott Kostyshak" <[email protected]> het volgende: > > On Sat, Aug 3, 2013 at 6:41 AM, Kornel Benko <[email protected]> wrote: > > Am Samstag, 3. August 2013 um 12:33:56, schrieb Kornel Benko > > <[email protected]> > >> Here the possible patch: > > > >> > > > >> The file "revertedTests" should contain the testnames, no blanks, each > >> testname in a separate line. > > > >> (E.g. > > > >> export/doc/Math_pdf > > > >> export/doc/he/Tutorial_pdf5 > > > >> ... > > > >> ) > > > >> > > > >> Kornel > > > > > > > > Sorry, this one is better. > > I like this. > > Note that the inverted test becomes: > INVERTED-TEST_FAILURE-IS-GOOD_export/doc/attic/DocStyle_pdf > instead of what we had mentioned above: > export/doc/attic/INVERTED-TEST_FAILURE-IS-GOOD_DocStyle_pdf > > I slightly prefer the new name (how it is with your patch) because you > can run both tests (even though there will be only one) by using the > original name. > > I don't agree with the name "revertedTests" (and likewise the variable > names). Using as a reverence > http://en.wiktionary.org/wiki/revert > I think that you are intending to use it with the following meaning: > "(transitive, now rare) To turn back, or turn to the contrary; to reverse." > As noted, this definition is rare. > I think most people running the tests will think of git (in fact, git > is mentioned on the wiktionary page) or of this definition: > "To cause to return to a former condition." > > Do you prefer "revert" to "invert"? > > Everything else looks good to me. >
Having tests that are known to fail is quite normal in a test suite. Why do we have to mangle these tests with those ugly verbose lengthy names? "Failure is good" makes not much sense. Failure is bad, but sometimes it is expected. I wouldn't mark them as good. Doesn't ctest have the option to add known failures ? We probably need to switch to another test framework anyway if we are going to introduce tests that look more like unit tests. Vincent
