Citing old or new papers of same work?

You might cite both by saying "Idea X, originally developed in [1] and extended in [2], shows that ..."

If you know why do you need a citation, it is usually easy to determine which papers to cite.

• If you want to give credit for a contribution, you cite the original paper.
• In computer science, journal papers usually supersede earlier conference versions of the same paper. If a journal version of the paper exists, you cite it instead of the conference version, because it is now the original paper.
• Subsequent papers by the same authors often contain new contributions to the topic. If the contributions are relevant to your paper, you cite those papers for the contributions.
• If you want to tell the reader what the state of the art is, you cite the most comprehensive papers on the topic. This may include survey articles or papers describing version 2.0 of the result, even if the specific contributions in 2.0 are not relevant to your paper.
• There are obviously other reasons for citing a paper. For example, you may want to cite historically important papers in a literature survey, even if the results are no longer relevant.

So, in general, you cite (the journal version of) the original paper for the contribution, and give the reader some relevant pointers to the state of the art. If the topic is central to your paper, you may also want to cite the intermediate papers for their specific contributions.