On Monday 13 March 2006 20:08, Adam Kennedy wrote: > I know of at least a few developers that would consider a report to be > bogus if one of their dependencies cannot install. Now personally, I > consider that a failure on the part of BOTH of them. > > One for failing, and the other for adding a compulsory dependency on > something that fails.
Bogus to me is "cannot install a dependency because, for all I can tell, the tester has a space somewhere in a directory name somewhere in the path to Perl and the tester's operating system and command shell just can't cope with this." That's well beyond something I can fix. Maybe it's not noise, in the sense that people in that situation might find it valuable information, but I don't find any value in getting test results from any system so broken that it cannot even begin to install dependencies. Of course, if I make screw up and somehow replace my Makefile.PL/Build.PL with the contents of a vim swap file and there's really nothing to run, the test report will be just as sensible -- so the distribution author marking a single test report as bogus isn't that interesting. Maybe having several separate developers flag a particular testing box as suspicious is valuable though. -- c
