Roman Yakovenko wrote: > I finally had time to investigate failures in pygccxml unittests. I will > list all changes I saw. > > Code, generated XML for both versions and essential differences could be > found in > http://pygccxml.svn.sourceforge.net/viewvc/pygccxml/pygccxml_dev/gccxml-0.9-upgrade/
Thanks! Do you mind if I link to this off the gccxml.org News page? > 1. Numerics is now reported with suffixes. For example instead 7, gccxml > 0.9 will report 7u or 7ul. See file const_variables.* files > 2. Demangled string has changed. For example see > http://pygccxml.svn.sourceforge.net/viewvc/pygccxml/pygccxml_dev/gccxml-0.9-upgrade/demangled.h.diff?revision=1118&view=markup > <http://pygccxml.svn.sourceforge.net/viewvc/pygccxml/pygccxml_dev/gccxml-0.9-upgrade/demangled.h.diff?revision=1118&view=markup> > file > 3. Type for pointers to member variable ( offset ) has changed. > 4. If struct doesn't contains user defined constructors, than new xml > will not contain them too. The previous version of GCC-XML generated > default & copy constructors > 5. Function default values contain "right values". > > I guess 1,2 & 4 are not bug and I need to update the code. I am not sure > about 3. I didn't take a look on new source code. These are all due to changes in GCC itself. Point #3 is a bit annoying but I do not see any reason that this was a wrong decision by the GCC folks. There is no such thing as a member-data type or reference-to-member type...there is only a pointer-to-member type. -Brad _______________________________________________ gccxml mailing list [email protected] http://www.gccxml.org/mailman/listinfo/gccxml
