Re: [boost] [chrono] New Features + Bug fixes + Updateddocumentation

Subject: Re: [boost] [chrono] New Features + Bug fixes + Updateddocumentation
From: Howard Hinnant (howard.hinnant_at_[hidden])
Date: 2009-12-13 12:55:37

On Dec 13, 2009, at 12:50 PM, vicente.botet wrote:

> Hi,
> ----- Original Message ----- 
> From: "Howard Hinnant" <howard.hinnant_at_[hidden]>
> To: <boost_at_[hidden]>
> Sent: Saturday, December 12, 2009 2:40 PM
> Subject: Re: [boost] [chrono] New Features + Bug fixes + Updateddocumentation
> 
> 
>> 
>> On Dec 12, 2009, at 5:22 AM, vicente.botet wrote:
>> 
>>> I have also installed MSVC and I have an issue with the ratio_fail_test1.cpp test. MSVC report a warning instead of an error. 
>>> 
>>> I have added 
>>> 
>>>       [ compile-fail ratio_fail_test1.cpp : <warnings-as-errors>on ]
>>> 
>>> and all is right.
>>> 
>>> Is there another way to make MSVC report an error when there is an integral constant overflow?
>> 
>> The intent is to have ll_mul cause an error with the C++0x feature "static_assert" instead of relying on the compiler to do it.  This line:
>> 
>>  static char test[a_x <= max / a_y];
>> 
>> is meant to fail at compile time if there is an overflow during multiplication.  I suggest you insert the boost macro version of static_assert (I'm not sure what the syntax is).  It has probably already been tuned to work well on MSVC.  If you do this throughout, everywhere I've got a commented out static_assert, and arrays of char, I think you'll get what you're looking for.  Sorry, I should've done this in the first place.
>> 
> 
> I have replaced the use of static_assert by BOOST_STATIC_ASSERT and added a conditional compilation on BOOST_CHRONO_USES_STATIC_ASSERT and now this works as expected.
> 
> Do you think it is worth removing the dependency to Boost.StaticAssert?

No, I think use of boost's static_assert is fine.  Over time that will migrate to the C++0X static_assert without this code needing to change.

-Howard