On Wed, May 25, 2022 at 2:34 AM Dave Chinner <david@xxxxxxxxxxxxx> wrote: > > On Tue, May 24, 2022 at 05:17:44PM +0300, Amir Goldstein wrote: > > On Tue, May 24, 2022 at 4:24 PM Dave Chinner <david@xxxxxxxxxxxxx> wrote: > > > On Tue, May 24, 2022 at 03:55:47PM +0300, Amir Goldstein wrote: > > The work you did is huge and impressive and the review is not easy. > > I disagree - it's not huge or impressive, it's just 4 hours of > *basic grunt work*. It's not difficult, it's not complex, it's just > time consuming. *Anyone* can do this. > > The problem fstests has is *nobody* is doing these sorts of > maintenance tasks. We keep adding more tests and with them mountains > of technical debt, yet nobody wants to take any responsibility for > addressing the technical debt. > Well said. And it is a sad reality that xfs maintainers are the ones having to clean up the mess. I'd imagine the xfs maintainer's pile is large enough as it is... It doesn't have to be this way. I have said it in LSFMM in "Maintainers don't scale" session lead by Josef (who stood in for Darrick) - Tech companies have many filesystem developers on the payroll. They also contribute resources to KernelCI and LTP. I think it is a matter of communicating to our managers that full/part time fstests developers on the payroll are a good way to meet their goals. Coming from fstests, when I first started working with LTP (for fanotify tests) I was amazed by the different experience: - Several LTP developers constantly working to improve the infrastructure - LTP developers reviewing merging and later improving the tests that you post - At some point, I saw a statement encouraging kernel developers to post reproducer in any form they see fit and the LTP developers will make it into an LTP test Good will of developers and weekend projects can only get us so far. If we want fstests to scale as a project, we need to nudge our managers to allocate headcount for dedicated fstests work. For that matter, I also consider the work on "fstests runners'' (i.e. fstests-bld and kdevops) as "fstests work", because making fstest more accessible to a wide audience will encourage more people to improve fstests itself. Thanks, Amir.
