On Tue, Nov 19, 2024 at 11:54:12AM -0800, Eduard Zingerman wrote: > On Tue, 2024-11-19 at 16:49 -0300, Arnaldo Carvalho de Melo wrote: > [...] > > > Lgtm. > > > Would it make sense to move this patch to the end of the series? > > > In case someone does a bisect and runs the tests to find some regression. > > > Humm? > > So you think it should be introduced only when it passes? I.e. when the > > problem is fixed? > Right. > > My practice so far has been to reproduce the problem manually, write a > > test, show that it detects the problem, fix, then show that the > > regression test shows that the problem is not present anymore. > Yes, the downside would be that anyone trying out the fix would > need to do some rebase to try the test w/o fix. > > I see your point about a bisection when running in the cset that > > introduces the test case and in all before the fix is added will fail, > > confusing the bisector or not allowing it to be automated :-\ > > So, yeah, probably, for automated bisection we should move it to after > > the fix, when finishing the devel cycle, which is now, will do. > Fwiw, that's what folks enforce for bpf selftests ¯\_(ツ)_/¯. Makes sense, automated bisection trumps old practices :-) - Arnaldo
