[cfe-dev] Proposal: -Wshadow-field flag

[Reid Kleckner]

On Tue, Apr 21, 2015 at 3:08 PM, Ehsan Akhgari <ehsan.akhgari at gmail.com>
wrote:

> On Tue, Apr 21, 2015 at 5:41 PM, David Blaikie <dblaikie at gmail.com> wrote:
>
>> FWIW I'd be a bit concerned that this warning, while it might have a good
>> false positive rate on a codebase with the particular idioms you've
>> described, would have a false positive rate a bit too high for a normal
>> diagnostic bar.
>>
>> Only one way to find out, though.
>>
>
> That is a good point.
>
> FWIW, I don't think we should turn this warning on as part of -Wall, maybe
> as part of -Wextra?  I'm not at all familiar with what the usual process
> for turning on warnings by default looks like, so I would appreciate if
> someone can point me in the right direction there.
>

Confusingly, "on by default" and "enabled by -Wall" are distinct in Clang,
but I don't find the distinction is not very useful.

I would suggest adding the warning under -Wshadow-field, and make it a
subgroup of -Wshadow. This way, -Wshadow users will get new warnings that
they probably wanted anyway. -Wshadow is not part of -Wall, so -Wall won't
enable it. Mark each new diagnostic as DefaultIgnore in
DiagnosticSemaKinds.td, so that it will not be on by default.

Once it's in, we can try it out, and if it has a really low false-positive
rate, then we can discuss enabling it by default and/or putting it in -Wall.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20150421/39f296b3/attachment.html>