[PATCH] Re-enabling another warning

David Blaikie dblaikie at gmail.com
Sat Jul 27 11:59:55 PDT 2013


On Sat, Jul 27, 2013 at 11:48 AM, Aaron Ballman <aaron at aaronballman.com> wrote:
> This patch fixes a few cases where a default destructor would not be
> generated due to a member variable having an inaccessible default
> destructor.  While the original code was correct, the updated code is
> a bit more clear.  Assuming that this new code is acceptable, this
> also re-enables the C4624 MSVC warning (''derived class' : destructor
> could not be generated because a base class destructor is
> inaccessible').

That code change does seem a bit strange/questionable/unnecessary.
While I'm OK enabling warnings in MSVC that have some substantial
overlap with Clang warnings to help developers working with MSVC keep
their work Clang-clean, generally we've taken the line that if a
warning isn't viable for Clang, it's not worth turning on in some
other compiler for LLVM. Does this warning catch useful bugs? (Why
would you need a warning for cases when a dtor couldn't be generated
when you'd get an error if/when you actually call that dtor? (because
that latter warning is hard to understand? That would be unfortunate))

I wonder why making an explicit dtor works? If the base class dtor is
inaccessible, wouldn't it still be inaccessible even in an explicit
dtor?

(& can we workaround this some other way? What happens if the default
ctor uses LLVM_DELETED_FUNCTION instead of an empty definition, does
that suppress the dtor warning? I doubt it, but figured I'd throw that
out there...)



More information about the llvm-commits mailing list