Messages in this thread | | | Date | Wed, 19 Apr 2017 16:08:04 +0200 | From | Peter Zijlstra <> | Subject | Re: [PATCH 2/3] jump_label: Provide static_key_slow_inc_nohp() |
| |
On Wed, Apr 19, 2017 at 08:49:11AM -0400, Steven Rostedt wrote: > I agree with Thomas. Removing the comment because a > "lockdep_assert_held()" exists at the top of the code, assumes someone > that is about to use that function did more that read the kerneldoc and > actually looked at the code.
No, it doesn't assume anything. You can use the function without reading _any_ comments. It will explode at runtime (preferably on the machine of the guy who wrote it -- who _SHOULD_ have lockdep enabled, but certainly on other developer's machines and test-bots, who will then loudly yell at said developer for not doing his job properly).
> If there's a kerneldoc to a function, than that header should contain > all the info that a developer needs to use that function.
Yeah, rainbows and unicorns are shiny. Also, I put kerneldoc (if I put it at all) at the definition site, not the declaration. So headers are useless.
In any case, I don't mind the extra line of comment. Don't really see the point of it either. What I am convinced of is that lockdep_assert_held() lines are far more useful than such comment lines.
| |