On Thu, Oct 20, 2016 at 5:25 PM, Eryu Guan <eguan@xxxxxxxxxx> wrote: > On Mon, Oct 17, 2016 at 10:01:19AM +0300, Amir Goldstein wrote: ... > > I'm still having concerns about losing test coverage by enabling "-i" by > default. How about adding an command line option to disable it? I could take the easy way and add a command line option to satisfy concerns, but we all know that 99% of the time, nobody is going to use it, so we will be just silencing out concerns instead of addressing them. > So at > least we could have a way to turn it off. Or is it completely impossible > to lose any test coverage? I would not argue that it is completely impossible to loose test coverage but I would argue that there is low probability of loosing *interesting* test coverage. Here is the argument I am making: A file reference leak can be anywhere, (e.g in EXDEV error path of clone file range ioctl) so we SHOULD have -i test coverage for as many APIs as possible with as many arguments as possible. OTOH, the difference between slowpath and fastpath of fdget()/fdput() for single/multi threaded process is quite invariant to the specific API, so I do not see the value in test coverage of fastpath to all APIs. Anyway, I'd be happy to add a command line option or env var if that is the preference of folks, but then I would really appreciate a suggestion for option name or var name, because it is going to be quite hard to find an intuitive name for this test variant. Thanks, Amir. -- To unsubscribe from this list: send the line "unsubscribe fstests" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
