bool to size_t warning

[Jonny Grant]

On 28/08/2019 01:07, Martin Sebor wrote:
> On 8/9/19 9:16 PM, Xi Ruoyao wrote:
>> On 2019-08-09 22:00 +0100, Jonny Grant wrote:
>>>
>>> On 09/08/2019 21:55, Maksim Fomin wrote:
>>>> ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
>>>> On Friday, August 9, 2019 8:32 PM, Jonny Grant <jg@jguk.org> wrote:
>>>>
>>>>> Hi
>>>>> Looks like the bool can't be converted to size_t
>>>>>
>>>>> Seems a shame it can't also convert to size_t
>>>>>
>>>>> What is strange in the example below, is I change it only a single
>>>>> conversion, or as follows, it compiles ok, so there must still be a
>>>>> conversion?
>>>>>
>>>>> Is this an issue worth reporting as a PR?
>>>>>
>>>>> size_t i = a;
>>>>> i += b;
>>>>>
>>>>> The C spec states |true|which expands to the integer
>>>>> constant|1|,|false|which expands to the integer constant|0|
>>>>>
>>>>> #include<cstddef>
>>>>>
>>>>> intmain()
>>>>> {
>>>>> boola = false;
>>>>> boolb = true;
>>>>> //size_t i = a;
>>>>> size_t i = a + b;
>>>>>
>>>>> returni;
>>>>> }
>>>>>
>>>>> #1 with x86-64 gcc (trunk)
>>>>> <source>: In function 'int main()':
>>>>>
>>>>> <source>:9:18: error: conversion to 'size_t' {aka 'long unsigned int'}
>>>>> from 'int' may change the sign of the result [-Werror=sign-conversion]
>>>>>
>>>>> Â Â Â Â Â  9 |Â Â Â Â  size_t i = a + b;
>>>>>
>>>>> Â Â Â Â Â Â Â  |Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â  ^
>>>>>
>>>>> cc1plus: some warnings being treated as errors
>>>>>
>>>>> Compiler returned: 1
>>>>>
>>>>> Please include my email in any replies
>>>>> Thanks, Jonny
>>>> 1. You mention C language but use C++ compiler: #include<cstddef> and
>>>> 'cc1plus'.
>>>>
>>>> 2. You can find answer in C++ language latest draft (
>>>> http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/n4820.pdf)
>>>> [expr.add], [expr.arith.conv], [conv.prom].
>>>>
>>>> TL;DR '+' promotes bools to signed integer which is not compatible with
>>>> size_t' {aka 'long unsigned int'}.
>>>>
>>>> You can use  size_t i = a + (size_t)b.
>>>>
>>>
>>> Fair enough. It is a shame the C++ spec didn't allow them to be treated
>>> as signed, or unsigned, they'll certainly never be negative anyway.
>>
>> "signed bool" does not make sense, I think :).
>>
>> The problem is that "-Wconversion" and "-Wsigned-conversion" really 
>> need some
>> improvement.
> 
> GCC does manage to avoid these warnings for constants that don't
> become negative.  It also has enough information to avoid them in
> cases when the promotion is from an unsigned type that's narrower
> than int, such as unsigned char or unsigned short.  But it doesn't
> consider the original type.  Clang implements the same
> -Wsign-conversion warning but it doesn't issue it for this case.
> I agree that it would make the GCC warning more useful if it
> handled these cases gracefully as well.
> 
> I would suggest to open a bug.
> 
> Martin
> 

Hi Martin

I created this PR

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91578

Thank you
Jonny