Messages in this thread | | | Date | Tue, 21 Jan 2014 16:43:19 -0800 | From | Randy Dunlap <> | Subject | Re: [PATCH 31/73] module: relocate module_init from init.h to module.h |
| |
On 01/21/2014 04:16 PM, Rusty Russell wrote: > Paul Gortmaker <paul.gortmaker@windriver.com> writes: >> Modular users will always be users of init functionality, but >> users of init functionality are not necessarily always modules. >> >> Hence any functionality like module_init and module_exit would >> be more at home in the module.h file. And module.h should >> explicitly include init.h to make the dependency clear. >> >> We've already done all the legwork needed to ensure that this >> move does not cause any build regressions due to implicit >> header file include assumptions about where module_init lives. >> >> Cc: Rusty Russell <rusty@rustcorp.com.au> >> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> > > Acked-by: Rusty Russell <rusty@rustcorp.com.au> > > Want to delete the extraneous semicolons, for bonus points? :) > >> +#define module_init(x) __initcall(x); > ... >> +#define module_exit(x) __exitcall(x);
There are still around 380 modules that use "module_init(x)" without a trailing semi-colon. That would make lots of build errors/warnings, I think. There was also a 2012 patch to fix those and change the module_init() macro.
What happened there?
ISTR (maybe incorrectly) that Linus once argued for leaving the trailing semi-colon there (perhaps 8 years ago). who knows?
-- ~Randy
| |