Re: [Quantlib-dev] QuantLib-1.8 | vc14 | AdaptiveRungeKutta testsuite error

[Ivan A. C. <iva...@gm...>]

Hello, Luigi.

The testsuite succeeded with your patch on “Version 14.0.25422.01 Update 3”:

1>------ Build started: Project: testsuite, Configuration: Release x64
------
…
1>  Testing observable settings...
1>  Testing adaptive Runge Kutta...
1>  Testing matrix exponential based on ode..
…
1>
1>  Test module "Master Test Suite" has passed with:
1>    646 test cases out of 646 passed
1>    1002058 assertions out of 1002058 passed
1>
========== Build: 1 succeeded, 0 failed, 1 up-to-date, 0 skipped ==========

So, supposedly, it is indeed the “Issue 3 Potential miscompilation of
code-calling functions that resemble std::min/std::max on floating point
values.” (https://msdn.microsoft.com/en-us/library/mt752379.aspx, Bugs that
are fixed in version 14.0.25425.01, released on 08/02/2016).

Personally, I wouldn’t recommend applying this patch to the master given
that std::max is used in many places in QuantLib, and it seems that this is
a VC compiler issue (I don’t recall having issues building 1.8 with GCC).

I will try to get VC updated and check if the issue disappears.

Thanks for looking into that.
On Sep 14, 2016 6:13 PM, "Luigi Ballabio" <lui...@gm...> wrote:

> Ivan,
>     please try the attached patch and check if it fixes the test failure.
>
> Thanks,
>     Luigi
>
>
>
> On Wed, Sep 14, 2016 at 10:49 AM Luigi Ballabio <lui...@gm...>
> wrote:
>
>> It seems so.  Anyway, I seem to have worked around the problem, but I'm
>> not in the office this morning. Stay tuned...
>>
>>
>> On Wed, Sep 14, 2016 at 10:45 AM Peter Caspers <pca...@gm...>
>> wrote:
>>
>>> Hi Luigi,
>>>
>>> not your fault, our subconscious mind presses these ok-buttons for us…
>>>
>>> Does it mean the problem only occurs in a specific version of VC14? In
>>> particular not in the latest?
>>>
>>> Regards
>>> Peter
>>>
>>> On 12 Sep 2016, at 15:27, Luigi Ballabio <lui...@gm...>
>>> wrote:
>>>
>>> Peter, Ivan,
>>>     I could reproduce the errors on a virtual machine running VC14.
>>> I'll try to find some time to work on this, too.
>>>
>>> Also, I'll share a life lesson I learned last week.  If you ever
>>> reproduce an error like the above, and then the compiler tells you that an
>>> update is available, for the love of all that's holy DON'T update.  If you
>>> do so, the error will disappear and your virtual machine will spend the
>>> best part of a Friday restoring the previous version of the compiler while
>>> you call yourself all sorts of names.  Just saying.
>>>
>>> Later,
>>>     Luigi
>>>
>>>
>>> On Fri, Sep 9, 2016 at 5:03 PM Ivan A. Cherkasov <
>>> iva...@gm...> wrote:
>>>
>>>> One more observation I wanted to share:
>>>>
>>>> with /fp:strict /fp:except compile flags (https://msdn.microsoft.com/
>>>> en-us/library/e7s85ffb.aspx) I’m getting build errors.
>>>>
>>>> command line:
>>>>
>>>> ^C:\TMP\QL18_VC2015_MK3\QUANTLIB-1.8\QL\EXPERIMENTAL\FINITEDIFFERENCES\
>>>> DYNPROGVPPINTRINSICVALUEENGINE.CPP
>>>> /c /IC:\BIN\BOOST_1_61_0 /I. /Zi /nologo /W3 /WX- /O2 /Ob2 /Ot /D
>>>> NDEBUG /D WIN32 /D _LIB /D _SCL_SECURE_NO_DEPRECATE /D
>>>> _CRT_SECURE_NO_DEPRECATE /D _MBCS /GF /Gm- /EHsc /MD /GS /Gy /fp:strict
>>>> /fp:except /Zc:wchar_t /Zc:forScope /Zc:inline /GR
>>>> /Fo".\BUILD\VC140\X64\RELEASE\\" /Fd".\BUILD\VC140\X64\RELEASE\QUANTLIB.PDB"
>>>> /Gd /TP C:\TMP\QL18_VC2015_MK3\QUANTLIB-1.8\QL\EXPERIMENTAL\
>>>> FINITEDIFFERENCES\DYNPROGVPPINTRINSICVALUEENGINE.CPP
>>>>
>>>> build errors:
>>>>
>>>> …
>>>>
>>>> 1>  dynprogvppintrinsicvalueengine.cpp
>>>> 1>C:\Program Files (x86)\Microsoft Visual Studio
>>>> 14.0\VC\include\xcomplex(646): error C2131: expression did not evaluate to
>>>> a constant (compiling source file ql\experimental\finitedifferences\
>>>> hestonrndcalculator.cpp)
>>>> 1>  C:\Program Files (x86)\Microsoft Visual Studio
>>>> 14.0\VC\include\xcomplex(646): note: failure was caused by an undefined
>>>> arithmetic operation (compiling source file ql\experimental\
>>>> finitedifferences\hestonrndcalculator.cpp)
>>>> 1>  ql\experimental\finitedifferences\hestonrndcalculator.cpp(110):
>>>> note: see reference to function template instantiation
>>>> 'std::complex<double> std::log<double>(const std::complex<double> &)' being
>>>> compiled
>>>>
>>>> …
>>>>
>>>> 1>C:\Program Files (x86)\Microsoft Visual Studio
>>>> 14.0\VC\include\xcomplex(232): error C2131: expression did not evaluate to
>>>> a constant (compiling source file ql\math\modifiedbessel.cpp)
>>>> 1>  C:\Program Files (x86)\Microsoft Visual Studio
>>>> 14.0\VC\include\xcomplex(232): note: failure was caused by an undefined
>>>> arithmetic operation (compiling source file ql\math\modifiedbessel.cpp)
>>>> 1>  C:\Program Files (x86)\Microsoft Visual Studio
>>>> 14.0\VC\include\xcomplex(253): note: see reference to function template
>>>> instantiation '_Ty std::_Fabs<double>(const std::complex<double> &,int *)'
>>>> being compiled
>>>> 1>          with
>>>>
>>>> ...
>>>>
>>>> and so on.
>>>>
>>>
>>>