Re: [boost] Additional Test Configurations

Subject: Re: [boost] Additional Test Configurations
From: Robert Ramey (ramey_at_[hidden])
Date: 2016-06-14 12:23:37

>>
>> I also see a whole load of linker errors which I don't understand at
>> present:
>>
>> gcc.link.dll
>> ..\..\..\bin.v2\libs\serialization\build\gcc-mingw-5.3.0s14\debug\libboost_serialization-mgw53-d-1_62.dll.a
>>
>>
>> ..\..\..\bin.v2\libs\serialization\build\gcc-mingw-5.3.0s14\debug\binary_iarchive.o:
>>
>> In function
>> `boost::serialization::singleton<boost::archive::detail::extra_detail::map<boost::archive::binary_iarchive>
>>
>>> ::get_mutable_instance()':
>> M:\data\boost\boost\libs\serialization\build/../../../boost/serialization/singleton.hpp:131:
>>
>> undefined reference to
>> `__imp__ZN5boost13serialization16singleton_module9is_lockedEv'
>> ..\..\..\bin.v2\libs\serialization\build\gcc-mingw-5.3.0s14\debug\binary_oarchive.o:
>>
>
> <snip>
> I'm not seeing this in any other platforms.

and I'm not seeing this on the test matrix even for mingw.

Another thing that intrigues me is the command line

"g++-4.4"  -ftemplate-depth-128 -Wno-c99-extensions -O0 -fno-inline 
-Wall -g -fPIC -m64 -Wno-unused-variable -Wno-long-long 
-ftemplate-depth-255 -DBOOST_ALL_NO_LIB=1 
-DBOOST_ARCHIVE_TEST=binary_archive.hpp -DBOOST_FILESYSTEM_DYN_LINK=1 
-DBOOST_LIB_DIAGNOSTIC=1 -DBOOST_SYSTEM_DYN_LINK=1 -D_GLIBCXX_DEBUG 
-D_STLP_DEBUG  -I".." -c -o 
"/var/boost/run/results/boost/bin.v2/libs/serialization/test/test_binary_binary_archive.test/gcc-4.4/dbg/test_binary.o" 
"../libs/serialization/test/test_binary.cpp"

I expect to see that -DBOOST_SERIALIZATION_DYN_LINK=1 and 
-DBOOST_WSERIALIZATION_DYN_LINK=1 as well.  It looks like the 
serialization libraries are being linked statically while the other 
libraries used in the test are being linked through DLLS.  My 
recollection is that this will link to both the static and shared C++ 
runtime libraries which is known to cause serious problems.

What am I missing here?

Also it's true that on my own system, I run develop branch for the 
libraries I'm testing against the master branch of all other libraries. 
This is different from the test matrix system which runs all packages 
from the develop or all packages from the master branch.  The fact that 
gcc passes on my system and fails on the test matrix might suggest that 
a bug or breaking change has been checked into the develop branch which 
is not being caught by some test.  I don't know.  I'm tempted to merge 
the serialization develop into the master just to see what happens - but 
I'm reluctant to go down that past.  My confidence in the test system 
has atrophied.

Robert Ramey

>
>>
>> HTH, John.
>>
>> _______________________________________________
>> Unsubscribe & other changes:
>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>
>
>
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost
>