Re: [boost] [Boost.Serialization] Crash in current master (1.68)

Subject: Re: [boost] [Boost.Serialization] Crash in current master (1.68)
From: Robert (r.firl_at_[hidden])
Date: 2018-07-02 19:03:49

On 7/2/2018 12:10 PM, degski via Boost wrote:
> On 2 July 2018 at 19:54, Robert via Boost <boost_at_[hidden]> wrote:
> 
>> Some of the team have updated to 15.7.4.  Because of my using the older
>> Microsoft version ...
>>
> 
> You can install the preview side-by-side with the stable version, you'd
> have BOBW (another advice from STL).

Yeah, I am not too excited about that on my primary desktop.  I might 
install the preview on a VM though.  Presently, the team continues to 
find newly released "features" that break stuff.  Hence, my now four 
decades (and growing) distrust...

> 
> This may not be known to news group members here.  But, the Windows Intel
>> C++ compiler is heavily dependent on the Microsoft supplied standard
>> library header files (e.g. from 2015: https://software.intel.com/en-
>> us/forums/intel-c-compiler/topic/596318).
> 
> 
> You said it, it's a compiler, not an implementation of a STL.
> 
> 
>> Similarly, in a Mac oriented post, there is a dependency on the gcc header
>> files (e.g. from 2013: https://software.intel.com/en-
>> us/forums/intel-c-compiler/topic/366102).
>>
> 
> Luckily I know nothing about Mac (in the same category as fuckbook.com,
> twitter (I don't own a phone) and smartphones in general (fat fingers, but
> my wife keeps asking me to debug her bloody Android phone, shit)).
> 
> 
>> I have already submitted an issue with Intel that the Visual Studio 15.7.4
>> headers cause Intel 18.0, Update 3 to produce thousands of spurious errors
>> on the std::variant.
>>
> 
> Yeah, I know from experience that they are very slow in reacting (if they
> react at all, which is usually not the case). F.e. the way they detect the
> compiler in tbb/mkl does not work with Clang/LLVM on windows, because they
> do the detection (if it should work) in the wrong order (they should check
> for VC first and Clang later, can't go wrong on linux/Mac), seems pretty
> simple, I filed it, but it is still as it used to be.

I have them fixing a couple other items too.  As far as speedily 
resolving something, that all depends on what the Intel group determines 
is important.  Tea leaf interpretation or flipping a coin 50 times, is 
about the same prediction accuracy.  :)

--Robert

> 
> degski
>