I have a paper (in computer science) which I have sent to a conference and I am planning to make a journal version which will contain some extra results. Is it acceptable to put the conference version to arXiv and later, when the journal version is done, add the new results to arXiv?
If the change from the conference version to the journal version is relatively minor (so the new results are unimportant), it's fine to add the new results to the arXiv.
If the change from the conference version to the journal version is substantial, post the conference version on the arXiv paper unchanged, and later post the journal version to the arXiv when you finish it. If you don't change the title from the conference paper to the journal paper, it might be a good idea to indicate in the arXiv paper titles that one is the conference version and the other is the journal version, so that people don't get confused.
This way, the papers on the arXiv reflect the published papers, so people who are trying to track things down in the literature won't get confused. You're shouldn't feel guilty about wasting resources, because you're not—the cost of hosting one more paper on the arXiv is tiny.
And you should consider whether the results you're adding to the conference paper are significant enough to deserve their own paper. Each paper should tell one "story". If the new results don't fit in the "story" of the paper, in my experience nobody will notice them, so you're better off writing a new paper containing them.