[cfe-dev] Doubt on a couple of warnings on deleted functions

[Richard Smith]

On Mon, Apr 21, 2014 at 2:00 PM, David Blaikie <dblaikie at gmail.com> wrote:

> On Mon, Apr 21, 2014 at 1:57 PM, Richard Smith <richard at metafoo.co.uk>
> wrote:
> >
> > On 21 Apr 2014 12:43, "David Blaikie" <dblaikie at gmail.com> wrote:
> >>
> >> On Mon, Apr 21, 2014 at 11:38 AM, Richard Smith <richard at metafoo.co.uk>
> >> wrote:
> >> > On Mon, Apr 21, 2014 at 7:47 AM, David Blaikie <dblaikie at gmail.com>
> >> > wrote:
> >> >>
> >> >> Yep
> >> >> Both of these sound like bugs and should be able to be fixed just by
> >> >> checking for a non-deleted definition. I'm not sure if there's a
> >> >> particular
> >> >> function to test that.
> >> >
> >> > I think the first warning is entirely appropriate. The point of this
> >> > warning
> >> > is that every external linkage function should be declared in a header
> >> > file;
> >>
> >> Sure, but how would you declare the deleted function in a header file
> >> separate from its delete definition?
> >>
> >> What you put in the header file is "void func() = delete;" and that's
> >> where the warning fires, isn't it?
> >
> > No, the warning fires if the first declaration of an externally visible
> > function is in the main source file.
>
> That doesn't seem to be consistent with what I'm seeing:
>
> $ cat deleted.h
> void func() = delete;
> $ cat deleted.cpp
> #include "deleted.h"
> $ clang++-tot deleted.cpp -Wmissing-prototypes -fsyntax-only -std=c++11
> In file included from deleted.cpp:1:
> ./deleted.h:1:6: warning: no previous prototype for function 'func'
> [-Wmissing-prototypes]
> void func() = delete;
>      ^
> 1 warning generated.


That's definitely a bug! =)

Looks like one issue is that isInlined() is incorrectly returning false for
deleted functions. An issue in the opposite direction is that there's no
'main source file' check (a forward declaration in the main source file
suppresses the warning, and really shouldn't).

(This warning also seems to be confused about whether it's looking for a C
function missing an explicit prototype and a C++ function missing a
previous declaration.)

>
> >> > it'd help if the warning suggested something like 'mark this function
> >> > 'static' if it's not intended to be used in other files' or similar.
> >> >>
> >> >> The unused variable one might crop up in defaulted functions too, in
> >> >> which
> >> >> case we'll want to check for a user provided definition.
> >> >
> >> > I agree on both counts. The same warning is visible here:
> >> >
> >> > struct S { S(const S&s); }; S::S(const S&s) = default;
> >> >
> >> >>
> >> >> On Apr 21, 2014 3:35 AM, "Nicola Gigante" <nicola.gigante at gmail.com>
> >> >> wrote:
> >> >>>
> >> >>> Hello.
> >> >>>
> >> >>> Consider this:
> >> >>>
> >> >>> $ cat example.cpp
> >> >>> void func(int x) = delete;
> >> >>> $ clang++ -std=c++11 -fsyntax-only -Wunused-parameter
> >> >>> -Wmissing-prototypes example.cpp
> >> >>> example.cpp:1:6: warning: no previous prototype for function 'func'
> >> >>> [-Wmissing-prototypes]
> >> >>> void func(int x) = delete;
> >> >>>      ^
> >> >>> example.cpp:1:15: warning: unused parameter 'x' [-Wunused-parameter]
> >> >>> void func(int x) = delete;
> >> >>>               ^
> >> >>> 2 warnings generated.
> >> >>>
> >> >>> The two warnings seems false positives:
> >> >>>
> >> >>> 1) why does it warn about a missing prototype? One can't add a
> >> >>> prototype
> >> >>> because
> >> >>> he'll get this error:
> >> >>>
> >> >>> error: deleted definition must be first declaration
> >> >>> void func(int x) = delete;
> >> >>>
> >> >>> So, the deleted definition is acting like a prototype itself,
> doesn't
> >> >>> it?
> >> >>> I think the warning
> >> >>> is wrong.
> >> >>>
> >> >>> 2) Does it make sense to warn about an unused parameter on a deleted
> >> >>> definition?
> >> >>> Of course is unused, but it can't be used anyway.
> >> >>> This force the programmer to avoid parameters name, like:
> >> >>> void func(int) = delete;
> >> >>> but this can reduce readability and it seems to me there's no reason
> >> >>> to
> >> >>> force it.
> >> >>>
> >> >>> Note that these warnings pop up only on free functions. Deleted
> member
> >> >>> functions
> >> >>> don't generate these warnings in these cases.
> >> >>>
> >> >>> Does it make sense to fix these issues?
> >> >>>
> >> >>> This mail is a part of a more general effort of mine to be able to
> >> >>> warning-less include
> >> >>> (with -Weverything, more or less) the LLVM headers that I use in my
> >> >>> project.
> >> >>> I think (and hope) more will follow.
> >> >>>
> >> >>> Thanks,
> >> >>> Nicola
> >> >>> _______________________________________________
> >> >>> cfe-dev mailing list
> >> >>> cfe-dev at cs.uiuc.edu
> >> >>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
> >> >>
> >> >>
> >> >> _______________________________________________
> >> >> cfe-dev mailing list
> >> >> cfe-dev at cs.uiuc.edu
> >> >> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
> >> >>
> >> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20140421/82d8c932/attachment.html>