[omniORB] Trouble building OmniOrb 4.1.4 on Linux

Igor Polonsky polonsig at yahoo.com
Wed Apr 28 21:25:02 BST 2010


Great thanks to Duncan and Tom for their most useful hints!

I've now managed to make and install the ORB.

It turned out that the main problems were caused by a few stupid errors on my part.  But that's not surprising for a guy who last did any Unix development 6.5 years ago.  Let me describe my adventures to you.

- First, I hadn't realized that the configure script would produce a config.log file. That's not mentioned in README.unix, unfortunately.  When Duncan asked me for a log file, I thought he meant a stdout redirect from the configure command.  But then I accidentally saw the "real" config.log, which is of course highly detailed and helpful.

- From the config.log file, I could see right away that a lot of the tests couldn't run because the C++ runtime shared library could not be found.  That was because I had failed to set LD_LIBRARY_PATH in my shell profile.  I had naively expected the configure script to set that variable automatically, since it knew where the gcc executable file lived.   Once I'd set that variable in my .profile, configure was able to do a much better job.  All the problems with numeric type sizes were now gone.

- But make was still failing.  The compiler (preprocessor) could not find some Python-related header files.  It turned out that the default Python installation on my box didn't have any includes.  So I pointed the configure script to a more complete installation of Python via the PYTHON parameter.  The help output implies that this parameter only sets the Python executable.  But somehow my setting it also helped the script to set the include directories right.  After that, it generated a set of files in the build directory that were good enough for make to do its job.

That said, I still see a few failed tests in config.log, which have to do with the lack of some floating point support functions and some socket-related structs.  I'm attaching the log file.  I'd really appreciate if you guys could take a look at it and let me know if any of these issues may create problems for me at run time.

Another question: I see quite a few compiler warning in the output produced by make.  I think that's Ok for a large and dynamic project.  But some of the warnings are of a kind that I personally try hard to avoid:

../../../../../omniORB-4.1.4/include/omniORB4/internal/initialiser.h:66:
warning: 'class omni::omniInitialiser' has virtual functions but non-virtual destructor
../../../../../omniORB-4.1.4/include/omniORB4/omniInterceptors.h:330:
warning: 'class omni::omniInterceptors::createThread_T::info_T' has virtual functions but non-virtual destructor
../../../../../omniORB-4.1.4/include/omniORB4/omniInterceptors.h:345:
warning: 'class omni::omniInterceptors::assignUpcallThread_T::info_T'has virtual functions but non-virtual destructor
../../../../../omniORB-4.1.4/src/lib/omniORB/orbcore/giopImpl10.cc:1681:
warning: 'class omni::omni_giopImpl10_initialiser' has virtual functions but non-virtual destructor

I'd appreciate your opinion on this issue.

Thanks,

Igor



      
-------------- next part --------------
A non-text attachment was scrubbed...
Name: config.zip
Type: application/x-zip-compressed
Size: 9478 bytes
Desc: not available
Url : http://www.omniorb-support.com/pipermail/omniorb-list/attachments/20100428/817e5195/config.bin


More information about the omniORB-list mailing list