Quoting Lee Jones (2021-02-12 14:37:39) > On Fri, 12 Feb 2021, Stephen Boyd wrote: > > > > > I'd like to enable it for only files under drivers/clk/ but it doesn't > > seem to work. I'm not asking to enable it at the toplevel Makefile. I'm > > asking to enable it for drivers/clk/ so nobody has to think about it now > > that you've done the hard work of getting the numbers in this directory > > down to zero or close to zero. > > I'm not sure which one of us is confused. Probably me, but ... > > Even if you could enable it per-subsystem, how would that help you? > > How can you ensure that contributors see any new W=1 warnings, but > Linus doesn't? When Linus conducts his build-tests during the merge > window, he is also going to build W=1 for drivers/clk. The assumption is contributors would have compiled the code they're sending, but that's obviously not always the case, so this assumption relies on developers running make. If they do run make then the hope is they would see the warnings now, without having to rely on them to know about passing W=1 to make, and fix them before sending code. If developers are ignoring build errors or warnings then we can't do anything anyway. > > All that's going to achieve is put you in the firing line. Ok. Is this prior experience? > > From my PoV W=1 builds should be enabled during the development phase > (i.e. contributor, auto-builder, maintainer). By the time patches get > make it into Mainline the review/testing stage is over and only the > default W=0 warnings are meaningful. > Alright maybe I don't understand and W=1 builds are noisy for the drivers/clk subdirectory even after applying these patches. Or it has some false positives that won't be fixed? Or a new compiler can cause new warnings to happen? I could see these things being a problem. I'm trying to see if we can make lives better for everyone by exposing the warnings by default in the drivers/clk/ directory now that there are supposedly none left. Shouldn't we tighten the screws now that we've cleaned them?
