Comment 8 for bug 910708

Yes, I agree with concerns about bad data. As for 'stopper' - I'm sure it's a stopper not for the development process, but for usage - i.e., when the tool is invoked in real-life environment, it leads to segfault.

As for the technical side - if I download 4 rpms attached to https://qa.mandriva.com/show_bug.cgi?id=64378 and run rpm only against them (e.g., 'rpm -qp --verify *rpm' - without installing them and then accessing rpmdb), then I also get a segfault.
Perhaps it is easier to perform debugging using this limited set of rpms then trying to bring rpmdb to a state that will cause problems. Though this can be a different issue, just with the same symptoms...