[cfe-dev] parentheses flag warning

[David Blaikie via cfe-dev]

On Mon, May 18, 2020 at 10:59 AM Billy Araujo <billyaraujo at gmail.com> wrote:

>
> Hi David,
>
> Actually the situation you present foo(x =5); is interesting because gcc
> 10 flags this while clang 10 doesn't:
>
> https://godbolt.org/z/NXUPrU
>
> warning: suggest parentheses around assignment used as truth value
>
>
Ah, well, missed opportunity for Clang, then! (bug/missing feature, I'd say
- though clang-to-clang backwards compatibility might make that hard to
roll out in clang at this point, not sure)


> @Roman. Thanks for the further explanation. That makes sense.
>
> Regards,
>
> Billy.
>
>
>
> On Mon, May 18, 2020 at 6:49 PM Roman Lebedev <lebedev.ri at gmail.com>
> wrote:
>
>> On Mon, May 18, 2020 at 8:44 PM Billy Araujo via cfe-dev
>> <cfe-dev at lists.llvm.org> wrote:
>> >
>> > Hi David,
>> >
>> > So I guess this is expected behaviour. Thanks.
>> >
>> > Regards,
>> >
>> > Billy.
>> >
>> >
>> > On Mon, May 18, 2020 at 6:39 PM David Blaikie <dblaikie at gmail.com>
>> wrote:
>> >>
>> >> GCC doesn't seem to warn on this either. On the basis of the original
>> warning I could see how this case could merit similar handling - but I
>> doubt we'd want to add that handling to the existing flag at this point
>> (due to the number of people over the decades that have used the warning
>> for the behavior it currently provides, not expecting the new/additional
>> behavior for the ternary operator or other bool-testing assignment
>> situations (eg: "void f(bool); f(x = 5);")
>> >>
>> >> On Mon, May 18, 2020 at 6:53 AM Billy Araujo via cfe-dev <
>> cfe-dev at lists.llvm.org> wrote:
>> >>>
>> >>> -Wparentheses
>> >>>
>> >>> warns for line 3 but not for the ternary expression in line 2.
>> >>>
>> >>> 1 static void foo(int a, int b, int x) {
>> >>> 2    x = (x = 10) ? a : b;
>> >>> 3    if (x = 10) { x = a; } else { x = b; }
>> >>> 4 }
>> >>>
>> >>> Is this a bug?
>> This seems pretty correct to me.
>>
>> You can't avoid braces on line 3, so those braces don't count as extra
>> braces
>> to silence the diagnostic, while the situation on the line 2 is opposite,
>> no braces are needed there, so the braces count as silencers.
>>
>> I.e. if line 2 *would* be diagnosed, i'd say that is a false-positive.
>>
>> >>> Regards,
>> >>>
>> >>> Billy.
>> Roman
>>
>> >>> _______________________________________________
>> >>> cfe-dev mailing list
>> >>> cfe-dev at lists.llvm.org
>> >>> https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>> >
>> > _______________________________________________
>> > cfe-dev mailing list
>> > cfe-dev at lists.llvm.org
>> > https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20200518/d3c94c36/attachment.html>