Re: Bad-sounding schur.cc compilation warnings

[Rik]

On 05/02/2016 12:40 PM, Daniel J Sebald wrote:
> On 05/02/2016 02:26 PM, Rik wrote:
>> On 05/02/2016 11:28 AM, address@hidden wrote:
>>> Subject:
>>> Bad-sounding schur.cc compilation warnings.
>>> From:
>>> Daniel J Sebald <address@hidden>
>>> Date:
>>> 05/02/2016 10:59 AM
>>>
>>> To:
>>> Maintainers List Octave <address@hidden>
>>>
>>> List-Post:
>>> <mailto:address@hidden>
>>> Content-Transfer-Encoding:
>>> 7bit
>>> Precedence:
>>> list
>>> MIME-Version:
>>> 1.0
>>> Message-ID:
>>> <address@hidden>
>>> Content-Type:
>>> text/plain; charset=utf-8; format=flowed
>>> Message:
>>> 3
>>>
>>>
>>> I saw the following warnings during compilation.  They look kind of
>>> troublesome:
>>>
>>> /home/sebald/octave/octave/octave/liboctave/numeric/schur.cc: In
>>> member function 'octave_idx_type schur<T>::init(const T&, const
>>> string&, bool) [with T = Matrix; octave_idx_type = int; std::string =
>>> std::basic_string<char>]':
>>> /home/sebald/octave/octave/octave/liboctave/numeric/schur.cc:176:19:
>>> warning: variable 'selector' might be clobbered by 'longjmp' or
>>> 'vfork' [-Wclobbered]
>>>    double_selector selector = 0;
>>>                    ^
>>> /home/sebald/octave/octave/octave/liboctave/numeric/schur.cc: In
>>> member function 'octave_idx_type schur<T>::init(const T&, const
>>> string&, bool) [with T = FloatMatrix; octave_idx_type = int;
>>> std::string = std::basic_string<char>]':
>>> /home/sebald/octave/octave/octave/liboctave/numeric/schur.cc:265:18:
>>> warning: variable 'selector' might be clobbered by 'longjmp' or
>>> 'vfork' [-Wclobbered]
>>>    float_selector selector = 0;
>>>                   ^
>>> /home/sebald/octave/octave/octave/liboctave/numeric/schur.cc: In
>>> member function 'octave_idx_type schur<T>::init(const T&, const
>>> string&, bool) [with T = ComplexMatrix; octave_idx_type = int;
>>> std::string = std::basic_string<char>]':
>>> /home/sebald/octave/octave/octave/liboctave/numeric/schur.cc:354:20:
>>> warning: variable 'selector' might be clobbered by 'longjmp' or
>>> 'vfork' [-Wclobbered]
>>>    complex_selector selector = 0;
>>>                     ^
>>> /home/sebald/octave/octave/octave/liboctave/numeric/schur.cc: In
>>> member function 'octave_idx_type schur<T>::init(const T&, const
>>> string&, bool) [with T = FloatComplexMatrix; octave_idx_type = int;
>>> std::string = std::basic_string<char>]':
>>> /home/sebald/octave/octave/octave/liboctave/numeric/schur.cc:463:26:
>>> warning: variable 'selector' might be clobbered by 'longjmp' or
>>> 'vfork' [-Wclobbered]
>>>    float_complex_selector selector = 0;
>>>
>>> Dan
>>
>> The warnings are ugly, but they aren't hazardous to Octave.  The wrapper
>> around Fortran function calls can potentially use longjmp to recover
>> from errors which is why you're getting the message.  If you look at the
>> actual C++ code it immediately returns after calling the Fortran
>> function and so any clobbered value would not get used.
>>
>> Most of the time we have used the volatile keyword as part of the
>> variable declaration just to quiet the compiler.  There is probably a
>> small performance penalty for doing so, but since we've done it
>> everywhere else, we could also do it here.  For what it's worth, I don't
>> get any such nasty warning with gcc-5.2.1.
>
> OK. I've version 4.8.4:
>
> gcc version 4.8.4 (Ubuntu 4.8.4-2ubuntu1~14.04.1)
>
> Mint/Unbuntu seems behind a bit in software versions, maybe with the
> intent of being more bug stable.
>
> Dan
>
Dan,

Can you try the attached cset?  It should suppress the warning message.

--Rik

schurv.cset
Description: Text document