Alekskey,
Seems like some recent changes brought MSVC to its knees -- http://tinyurl.com/3vmlv.
Fix is already committed. Tomorrow we should be all green again. My plan is for this release is to get maximum compatibility with the older compilers. In particular VC6.5 7.0, GCC 2.95, Intel 7.0, and CV8.3 Then after this release I can drop many of the multitude of workaround necessary for these. With regard to Code Warrior I was hoping to develop a fix for the debug mode compilation errors show up on you metacomm test. I didn't come up with anything satisfactory however. Do you have any idea why there is such a large difference between your cw-8.3 and the RSI's cw-8.3-release tests? I can only see the command line options in the failing build. The only option I can see which could be of relevance is the '-iso-template on'. If you are both using stock versions of the build system this same option should be set for the release build as well. Thanks for you interest (and hard work), Michael
Michael Stevens wrote:
With regard to Code Warrior I was hoping to develop a fix for the debug mode compilation errors show up on you metacomm test. I didn't come up with anything satisfactory however. Do you have any idea why there is such a large difference between your cw-8.3 and the RSI's cw-8.3-release tests?
No clue from me :-( But I'll try and run the uBLAS tests in debug mode tonight and see if anything comes up.
I can only see the command line options in the failing build. The only option I can see which could be of relevance is the '-iso-template on'. If you are both using stock versions of the build system this same option should be set for the release build as well.
My "cw-8.3-release-tools.jam" is an exact copy of "cw-8.3-tools.jam". Only the name changed to distinguish my runs as using the release BUILD. You might want to look at the fails on other libraries on my configuration for the command line options. And looking at them now I see one difference. The metacomm command lines have -DEXTERNAL. Don't know what that's for. --grafik
Rene Rivera wrote:
No clue from me :-( But I'll try and run the uBLAS tests in debug mode tonight and see if anything comes up.
I ran both release and debug tests for uBLAS and the debug versions also fail on my setup.
My "cw-8.3-release-tools.jam" is an exact copy of "cw-8.3-tools.jam". Only the name changed to distinguish my runs as using the release BUILD.
The debug and release command lines on my setup are, accordingly, for the concept test only: mwcc -c -lang c++ -DEXTERNAL -maxerrors 5 -warn on -maxwarnings 20 -nowraplines -g -O0 -inline off -cwd include -warn nounusedexpr,nounused -runtime dmd -DNOMINMAX -RTTI on -iso_templates on -I- "-I/usr/include" "-IC:\CVSROOTs\Boost\boost" -prefix "UseDLLPrefix.h" -o "C:\Boost\test\bin\boost\libs\numeric\ublas\concepts.test\cw\debug\threading-multi\concepts.obj" "..\..\..\libs\numeric\ublas\concepts.cpp" mwcc -c -lang c++ -DNDEBUG -DNDEBUG -DEXTERNAL -maxerrors 5 -warn on -maxwarnings 20 -nowraplines -opt full -inline auto -inline level=8 -cwd include -warn nounusedexpr,nounused -runtime dm -DNOMINMAX -RTTI on -iso_templates on -I- "-I/usr/include" "-IC:\CVSROOTs\Boost\boost" -prefix "UseDLLPrefix.h" -o "C:\Boost\test\bin\boost\libs\numeric\ublas\concepts.test\cw\release\threading-multi\concepts.obj" "..\..\..\libs\numeric\ublas\concepts.cpp" I also did a pre-process of both of those commands. The results of that are attached. Hmm, they are a bit large... Instead I'll send them to Micheal directly. -- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org - grafik/redshift-software.com - 102708583/icq
participants (6)
-
Michael Stevens -
Rene Rivera