On Mon, Aug 24, 2020 at 09:28:28AM -0400, Steven Rostedt wrote: > On Mon, 24 Aug 2020 08:26:13 +0000 > David Laight <David.Laight@xxxxxxxxxx> wrote: > > > Since the objective seems to be to ensure there are no > > calls to trace_printk() in the git tree, but to allow > > them in uncommitted sources. Why not use a config option > > and rely on rand-config builds to detect any 'accidental' > > commits? > > Because we don't want distros to ship with disallowing trace_printk, > where someone finds a bug, sends the config to a developer, who then > adds trace_printk() just to find that they can't use it. > > The point of avoiding a config was to keep the burden of having the > developer needing it and having to then modify the config given to them. > > Also, it would then prevent those developing modules from using > trace_printk() in their module if they build against one of these > kernels. > > Finally, there's debug code in the kernel that legitimately uses > trace_printk(), and those randconfigs will trigger on them. How making it make's option prevent some "smart" distros to achieve the same? AFAIU any compile-time knob will allow to build a kernel w/o a feature and you are against of such kernel builds in distros. Catch-22? -- With Best Regards, Andy Shevchenko
