From: Robert Ramey (ramey_at_[hidden])
Date: 2004-10-22 19:12:45
I rebuilt and reran the the tests "test_class_info_load" and
"test_class_info_save" which are explicitly designed to test this
functionality. I ran these test with XML archives. Do these tests pass for
you? How is this situation different than yours? I updated the test
sliightly so that the class version jumps from 2 to 4 - skipping over 3 so
that a confusion between serialization library version and class version
wouldn't be overlooked.
"Russell Hind" <rh_gmane_at_[hidden]> wrote in message
> Robert Ramey wrote:
> > There is a confusion here. Check the manual section titled "Class
> > Versioning"
> > There should be something like:
> > BOOST_CLASS_VERSION(Test_c, 99)
> > in your code. I don't remember what the maximum class version might be
> > it could be small like 255 for binary archives.
> I did ask about this in the previous e-mail. I thought non-primitive
> types were, by default, versioned, i.e. if no BOOST_CLASS_VERSION is
> specified, they have a version of 0. Is this not the case?
> Anyway, the code I posted still gives odd results if I set
> BOOST_CLASS_VERSION(Test_c, 0). Anything but 0 works, but zero gives
> the effects described previously, a 3 on the way back in, which is your
> archive version, not the class version.
> Again, specifying no version works for text archives (haven't checked
> binary) but not for XML.
> Unsubscribe & other changes:
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk