On Tue, Nov 06, 2007 at 10:40:30PM -0600, William Rowe wrote: > Joe Orton wrote: > >On Tue, Nov 06, 2007 at 12:50:44AM -0000, William Rowe wrote: > >>Author: wrowe > >>Date: Mon Nov 5 16:50:41 2007 > >>New Revision: 592215 > >> > >>URL: http://svn.apache.org/viewvc?rev=592215&view=rev > >>Log: > >>It is entirely pointless to have nonportable behaviors as examples > >>to end users of the library. Good point, however, for an @tip. > > > >That was a test case for a bug in the implementation of > >apr_dir_make_recursive (see r291339). You have certainly *made* it > >entirely pointless. > > If we are testing make_recursive, why are we we doing so in this context? > That particular implementation has some assumptions, right, I get that. > We have a place for implementation tests, it's within test/internal/, where > we removed the testucs module you objected to way back when.
Eh? The test case in question was for a bug in the implementation of a public function which happens to only trigger on some subset of supported platforms, like many other tests in the test suite. > It's a good solution; for trunk and moving forwards, lets adopt it. > > >Please revert this and explain what problem you were seeing with the > >test case. And please use commit log messages which describe the change > >you are making, rather than some abstract commentary. > > Message documented in STATUS. Nothing should be emitted in ./testall -v > that doesn't reflect an implementation oversight, an implementation flaw > or an outright bug in apr or it's OS. This particular case was not a flaw, > as I spelled out in apr_file_io.h's @warning. So the only problem you had with the test case was that it happened to call ABTS_NOT_IMPL? Um, OK. Please restore the test case already and remove the ABTS_NOT_IMPL call. joe
