Should I use full URLs or paths from the root of the application on my site?

I tried to originally use full 开发者_运维知识库paths to the various pages on my site because it was better for the SEO of my site, and thus was a business-driven requirement.

But now a bug has been discovered that occurs when users log in, and the full path causes the session to think they are not logged in. So they have to log in again for the auth mechanism to finally realize they are logged in.

Has anyone ever seen that before? What is a good solution? Ideally I'd need to keep my full url paths.

Thanks, Alex