On 3/5/24 6:49 PM, Jani Nikula wrote: > On Tue, 05 Mar 2024, Johannes Berg <johan...@sipsolutions.net> wrote: >> On Tue, 2024-03-05 at 12:34 +0200, Jani Nikula wrote: >>> We're seeing a lot of -Wmissing-prototypes warnings when building >>> user-mode Linux. For example, our CI hits things like [1], and it's just >>> extra noise in the build. >> >> Yeah ... >> >>> I started adding some prototypes, but I lost my way around the arch/ >>> subdirectory a bit. I'm not always quite sure what should or could be >>> included where, or whether a new header should be added instead, etc., >>> so it wasn't going as quickly as I'd hoped. >>> >>> Do you perhaps already have plans or, better yet, patches to address >>> this? Feels like it should be fairly quick for someone familiar with the >>> arch/um codebase. >> >> I think this series addressed most already? >> >> https://patchwork.ozlabs.org/project/linux-um/list/?series=393619 >> >> But I haven't checked myself yet. > > Oh, nice, it certainly helps! > > But after applying that in my branch, I do still get a bunch of > warnings.
Thanks for raising this topic! Indeed. To make it easier to review, the current series only addressed the -Wmissing-prototypes warnings that can be fixed straightforwardly (e.g. by turning into static functions or including missing headers). The remaining -Wmissing-prototypes warnings mainly fall into the following categories: - Symbols defined and used in os-Linux/; - Symbols defined in os-Linux/, but declared in kernel headers; - Symbols defined in individual files and not called explicitly; My plan was to fix the remaining warnings in followup series. Ultimately, I wish we can let cc treat -Wmissing-prototypes warnings as errors. Regards, Tiwei