Re: [boost] [predef] Using predef-check on 'develop' problem

Subject: Re: [boost] [predef] Using predef-check on 'develop' problem
From: Rene Rivera (grafikrobot_at_[hidden])
Date: 2015-06-16 21:56:11

On Tue, Jun 16, 2015 at 7:03 PM, Edward Diener <eldiener_at_[hidden]>
wrote:

> On 6/10/2015 12:36 AM, Rene Rivera wrote:
>
>> On Tue, Jun 9, 2015 at 10:34 PM, Rene Rivera <grafikrobot_at_[hidden]>
>> wrote:
>>
>>  On Tue, Jun 9, 2015 at 9:56 PM, Jessica Hamilton <
>>> jessica.l.hamilton_at_[hidden]> wrote:
>>>
>>>  On 9 June 2015 at 19:40, Rene Rivera <grafikrobot_at_[hidden]> wrote:
>>>>
>>>>> On Fri, Jun 5, 2015 at 9:42 PM, Edward Diener <eldiener_at_[hidden]
>>>>> >
>>>>> wrote:
>>>>>
>>>>>  I think I know what is wrong in the predef-check functionality. In my
>>>>>>
>>>>> VMD
>>>>
>>>>> jamfile the use of predef-check, for any given compile or run rule,
>>>>>>
>>>>> looks
>>>>
>>>>> like:
>>>>>>
>>>>>> [ predef-check "BOOST_COMP_GNUC >= 4.3" "BOOST_OS_QNX == 0" : :
>>>>>> <cxxflags>-std=c++0x ]
>>>>>>
>>>>>>
>>>>> Yes..
>>>>>
>>>>> What I am seeing is that except on QNX, where "BOOST_OS_QNX != 0", the
>>>>>
>>>>>> '-std=c++0x' is always being added as a C++ compiler flag to the
>>>>>>
>>>>> command
>>>>
>>>>> line.
>>>>>>
>>>>>> This initially suggests to me that you are treating more than one
>>>>>>
>>>>> quoted
>>>>
>>>>> predef definition as an OR gate rather than an AND gate. But note that
>>>>>>
>>>>> on
>>>>
>>>>> QNX, where where "BOOST_OS_QNX != 0" and "BOOST_COMP_GNUC >= 4.3" is
>>>>>>
>>>>> true,
>>>>
>>>>> the '-std=c++0x' is not being added. So your logic seems to be that as
>>>>>>
>>>>> you
>>>>
>>>>> go through multiple predef definitions once you hit a 'true' condition
>>>>>>
>>>>> you
>>>>
>>>>> choose the 'true' path as long as no 'false' conditions follow it,
>>>>>>
>>>>> else you
>>>>
>>>>> choose the 'false' path.
>>>>>>
>>>>>> Please see if you can fix this given this clue about how predef-check
>>>>>>
>>>>> is
>>>>
>>>>> working for the VMD regression tests on various platforms/compilers.
>>>>>>
>>>>>
>>>>>
>>>>> I don't know if it helped but I did a change to one of the tests I do
>>>>> to
>>>>> completely cover all the Venn variations of the binary and expression.
>>>>> I
>>>>> changed my test to this:
>>>>>
>>>>>      [ run check_value.cpp : : : <test-info>always_show_run_output
>>>>>          [ predef-check "BOOST_COMP_CLANG > 0" "BOOST_OS_LINUX == 0" :
>>>>> :
>>>>> <cxxflags>-DCHECK_VALUE=true : <cxxflags>-DCHECK_VALUE=false ] ]
>>>>>
>>>>
>>>> Out of curiosity, what happens if you remove the second <cxxflags>
>>>> variable, and use an #ifndef check, and run again? It's the only
>>>> difference I can notice between the two examples.
>>>>
>>>>
>>> OK.. Did that change. My local OSX test shows "CHECK_VALUE == 1" as
>>> expected and correctly. In a few minutes I'll post what the online tests
>>> show. But the expectation is that they should all show "CHECK_VALUE ==
>>> undefined".
>>>
>>>
>> And the online tests are now done. And indeed they all print out
>> "CHECK_VALUE == undefined".
>>
>
> Please look at the latest 'develop' results for SunOS. The value of
> BOOST_COMP_CLANG = 0 and yet your check_value test shows CHECK_VALUE == 1.
> Also the msvc-14.0 tests under Windows shows the same problem.

I'm looking :-) Such a painful problem to track down :-(

-- 
-- Rene Rivera
-- Grafik - Don't Assume Anything
-- Robot Dreams - http://robot-dreams.net
-- rrivera/acm.org (msn) - grafikrobot/aim,yahoo,skype,efnet,gmail