[LLVMdev] [RFC] C++11: 'virtual' and 'override'

[David Blaikie]

On Sun, Mar 2, 2014 at 8:19 PM, Craig Topper <craig.topper at gmail.com> wrote:
> While doing the conversion of LLVM_OVERRIDE to 'override' last night, I
> noticed that the code base is rather inconsistent on whether the 'virtual'
> keyword is also used when 'override' is used.
>
> Should we have a coding standard for this? What's the preferred direction
> here? Seems not having 'virtual' is less overall text, but not sure how
> others feel.

My vote: omit virtual if override is used.

(legitimate counterargument: harder to skim/match/read whether a
function is virtual when it's not specified and "override" appears
much later in the declaration)

> Related, should we require use of 'override' when methods override a base
> class method?

My vote: require override.

> I have clang-tidy checks for both though haven't implemented fixits for them
> yet.

Cool. There has also been a suggestion that Clang could warn about
omitted override if at least one other member function in the same
class is marked override, which could get us a lot of value built into
the compiler (but not 'all the way', so a clang-tidy check would still
be appropriate).