[webkit-dev] Unused parameter warnings / errors / warning fixes

Filip Pizlo fpizlo at apple.com
Mon Jan 26 10:11:25 PST 2015

> On Jan 26, 2015, at 9:28 AM, Darin Adler <darin at apple.com> wrote:
> I have found serious bugs in WebKit with unused parameter warnings; I found some when first turning on the warning. I don’t have a specific story to tell, but I remember there was more than one case.
> So far it’s been impractical to turn on this warning for our Objective-C code, since we can’t simply omit argument names, and that’s why it’s not on in the Mac versions of those frameworks.
> And I do agree that it can be inconvenient to deal with these warnings in heavily #ifdef’d code. I think there are some good strategies for avoiding that, and I’d like to talk about specific irritating cases so I can propose the strategy I’d push for. Generally the strategy is to push more configuration specifics to the edges rather than spreading them through multiple functions.

A strategy that works well for me for reducing the total number of #ifdefs is to put them in the implementation of a method rather than around the whole method. Of course this strategy is ultimately unavoidable - if you have any method that is only declared on some configuration then somewhere you will have a conditionalized call to that method. So I actually do the opposite of pushing the configuration conditionals to the edges - I put them into the innards. I believe that this reduces both the total number of #ifdefs and the total number of places where unused parameter situations arise. My experience here may be because this is the best way to deal with things like CPU and OS; I don't know what it's like for feature flags. 

Either way though, you eventually have some code in some method body that will be conditionalized. This leads to two cases of compiler spoonfeeding:

1) unused parameters

2) unreachable code

We have a good way of dealing with (2) - UNREACHABLE_FOR_PLATFORM. This tells the compiler that you won't return but deactivates the noreturn attribute warning. That's probably as good as it gets for unreachable code - even if the noreturn warning wasn't enabled we would still want to make sure that presumed-unreachable code aborts if called. 

For unused parameters I don't know of a case where it caught a bug but I end up dealing with build failures - either on EWS or after commit - due to the warning. 

> — Darin
> _______________________________________________
> webkit-dev mailing list
> webkit-dev at lists.webkit.org
> https://lists.webkit.org/mailman/listinfo/webkit-dev

More information about the webkit-dev mailing list