Hi Robert,
Robert Osfield wrote:
Hi Alberto,
On Mon, May 18, 2009 at 8:02 PM, Alberto Luaces <[email protected]> wrote:
Hi,
El Lunes 18 Mayo 2009ES 20:24:18 Robert Osfield escribió:
nm OpenSceneGraph/lib/libosg.so -D --demangle | awk '{print $2, $3,
$4, $5, $6, $7}' > 2.8.0.nm
somebody correct me if I'm wrong, but I think the safest thing would be to
compare the mangled symbols, that are what the operating system and the linker
work with, after all.
In this instance I compiled the source libraries from source using the
same compiler and settings on the same day, so there isn't too much
scope for naming managing to change. I was concerned about one
particular change introduced today and it does look it's not an issue
from the nm tests I've just done.
As the Vec?b and BoundingSphere classes are pure inline classes
implemented in a header file (i.e. when included they are copied
verbatim into the including file), they do not show up in the export
list of a DLL/DSO and are therefore not caught by a nm/dumpbin test.
Paul Obermeier
As a general note, you are correct, one should compare the results
without demangling them.
I'm open to suggestion for a more formalized system for checking for
ABI breakages, gettings something for just unices would probably be
sufficient as it's not something we need to run everyday and have all
users testing it. It's sufficient for just maintainers like myself to
occasionally do double checks.
Robert.
_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
