Re: [glibc] malloc: Fix mallinfo deprecation declaration

Re: [glibc] malloc: Fix mallinfo deprecation declaration

[Florian Weimer]

* Adhemerval Zanella via Glibc-cvs:

> +  /* The test below covers the deprecated mallinfo function.  */
> +  DIAG_PUSH_NEEDS_COMMENT;
> +  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");

This should be 10 or 11, not 4.9.  The ignored diagnostics is required
for current compilers, and will always be.

Re: [glibc] malloc: Fix mallinfo deprecation declaration

[Adhemerval Zanella]

On 31/08/2020 15:28, Florian Weimer wrote:
> * Adhemerval Zanella via Glibc-cvs:
> 
>> +  /* The test below covers the deprecated mallinfo function.  */
>> +  DIAG_PUSH_NEEDS_COMMENT;
>> +  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
> 
> This should be 10 or 11, not 4.9.  The ignored diagnostics is required
> for current compilers, and will always be.
> 

I saw the deprecated warning on gcc 9.2.1 and thus I used the gcc version
used on some other usages of the same ignore:

  posix/tst-dir.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
  posix/tst-dir.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
  signal/tst-sigset2.c:DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
  signal/tst-sigsimple.c:DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
  stdio-common/tst-printfsz.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
  sysdeps/pthread/tst-cancel4.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
  time/tst-ftime.c:      DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");

This is essentially for __attribute_deprecated__, I am failing to see why this
will only happen with gcc 10 or 11.

Re: [glibc] malloc: Fix mallinfo deprecation declaration

[Florian Weimer]

* Adhemerval Zanella:

> On 31/08/2020 15:28, Florian Weimer wrote:
>> * Adhemerval Zanella via Glibc-cvs:
>> 
>>> +  /* The test below covers the deprecated mallinfo function.  */
>>> +  DIAG_PUSH_NEEDS_COMMENT;
>>> +  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>> 
>> This should be 10 or 11, not 4.9.  The ignored diagnostics is required
>> for current compilers, and will always be.
>> 
>
> I saw the deprecated warning on gcc 9.2.1 and thus I used the gcc version
> used on some other usages of the same ignore:
>
>   posix/tst-dir.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>   posix/tst-dir.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>   signal/tst-sigset2.c:DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>   signal/tst-sigsimple.c:DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>   stdio-common/tst-printfsz.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>   sysdeps/pthread/tst-cancel4.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>   time/tst-ftime.c:      DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>
> This is essentially for __attribute_deprecated__, I am failing to see why this
> will only happen with gcc 10 or 11.

The is supposed to refer to the last compiler version that was
verified to need the pragma.  The idea is to revisit those directives
periodically and remove them once the compilers that need them are no
longer supported for building glibc.

That does not apply to any of these cases.

Re: [glibc] malloc: Fix mallinfo deprecation declaration

[Adhemerval Zanella]

On 31/08/2020 16:16, Florian Weimer wrote:
> * Adhemerval Zanella:
> 
>> On 31/08/2020 15:28, Florian Weimer wrote:
>>> * Adhemerval Zanella via Glibc-cvs:
>>>
>>>> +  /* The test below covers the deprecated mallinfo function.  */
>>>> +  DIAG_PUSH_NEEDS_COMMENT;
>>>> +  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>>
>>> This should be 10 or 11, not 4.9.  The ignored diagnostics is required
>>> for current compilers, and will always be.
>>>
>>
>> I saw the deprecated warning on gcc 9.2.1 and thus I used the gcc version
>> used on some other usages of the same ignore:
>>
>>   posix/tst-dir.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>   posix/tst-dir.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>   signal/tst-sigset2.c:DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>   signal/tst-sigsimple.c:DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>   stdio-common/tst-printfsz.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>   sysdeps/pthread/tst-cancel4.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>   time/tst-ftime.c:      DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>
>> This is essentially for __attribute_deprecated__, I am failing to see why this
>> will only happen with gcc 10 or 11.
> 
> The is supposed to refer to the last compiler version that was
> verified to need the pragma.  The idea is to revisit those directives
> periodically and remove them once the compilers that need them are no
> longer supported for building glibc.
> 
> That does not apply to any of these cases.
> 

Yes and the deprecated warning also does really fit in the category you
described since it is defined by glibc itself.  I give you that '4.9' indeed
is not the best choise (although not wrong itself), but for this particular
case I think using the minimum support version version gcc required to 
build itself make more sense.

Maybe a best alternative is to define the minimum compiler version on
libc-diag.h and use it on "-Wdeprecated-declarations" instead.  I guess the
SELinux deprecated-declarations usage which are constraint for gcc 10 are
not entirely correct as well.

Re: [glibc] malloc: Fix mallinfo deprecation declaration

[Adhemerval Zanella]

On 31/08/2020 16:50, Adhemerval Zanella wrote:
> 
> 
> On 31/08/2020 16:16, Florian Weimer wrote:
>> * Adhemerval Zanella:
>>
>>> On 31/08/2020 15:28, Florian Weimer wrote:
>>>> * Adhemerval Zanella via Glibc-cvs:
>>>>
>>>>> +  /* The test below covers the deprecated mallinfo function.  */
>>>>> +  DIAG_PUSH_NEEDS_COMMENT;
>>>>> +  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>>>
>>>> This should be 10 or 11, not 4.9.  The ignored diagnostics is required
>>>> for current compilers, and will always be.
>>>>
>>>
>>> I saw the deprecated warning on gcc 9.2.1 and thus I used the gcc version
>>> used on some other usages of the same ignore:
>>>
>>>   posix/tst-dir.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>>   posix/tst-dir.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>>   signal/tst-sigset2.c:DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>>   signal/tst-sigsimple.c:DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>>   stdio-common/tst-printfsz.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>>   sysdeps/pthread/tst-cancel4.c:  DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>>   time/tst-ftime.c:      DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wdeprecated-declarations");
>>>
>>> This is essentially for __attribute_deprecated__, I am failing to see why this
>>> will only happen with gcc 10 or 11.
>>
>> The is supposed to refer to the last compiler version that was
>> verified to need the pragma.  The idea is to revisit those directives
>> periodically and remove them once the compilers that need them are no
>> longer supported for building glibc.
>>
>> That does not apply to any of these cases.
>>
> 
> Yes and the deprecated warning also does really fit in the category you
> described since it is defined by glibc itself.  I give you that '4.9' indeed
> is not the best choise (although not wrong itself), but for this particular
> case I think using the minimum support version version gcc required to 
> build itself make more sense.

I should revise what I wrote more often... I mean that 'using the minimum
supported gcc version to build glibc itself makes more sense'.

> 
> Maybe a best alternative is to define the minimum compiler version on
> libc-diag.h and use it on "-Wdeprecated-declarations" instead.  I guess the
> SELinux deprecated-declarations usage which are constraint for gcc 10 are
> not entirely correct as well.
>