Kris McDaniel and Ben Bradley have a great article, 'Desires' (forthcoming in Mind), which argues that desire is a three-place relation between a person and two propositions: the object of the desire, and the condition on its applicability. (Though for ordinary unconditional desires, this is a mere technicality: the condition is the trivial proposition that any old thing is the case.)
This is motivated by the thought that some desires are neither satisfied nor frustrated, but simply cancelled (or inapplicable). Consider a child's desire for Santa to have a nice Christmas, or my desire to have an icecream later if I still feel like it. Even if these things never come about, it doesn't necessarily seem as though my past desire was thwarted. (Ask my earlier self: "Do you desire that, even if your future self no longer feels like it, you still get the icecream anyway?" I will answer, "of course not!") The desires come with certain presuppositions: that Santa exists, or that I will continue to want icecream, etc. When those conditions fail, the desire simply no longer applies.
McDaniel and Bradley compare their view to Strawson's suggestion about non-referring expressions: "the question whether his statement ['the king of France is wise'] was true or false simply didn't arise, because there is no such person as the king of France." So it goes, I suppose, for the person who desires that the king of France be wise. The question whether their desire is fulfilled or frustrated simply doesn't arise. To further illustrate, the authors suggest that conditional desires are like conditional bets: if the condition isn't met, you neither win nor lose; instead, the bet is off.
How should a utilitarian take cancelled desires to figure in the calculus? The obvious option is that cancelled desires have no normative significance, whereas fulfilled desires have positive value, and thwarted desires have disvalue. So if we can find a desire the thwarting of which doesn't matter in the slightest, that might be taken to show that it's a conditional desire which has been cancelled.
One thing I disagree with: the authors claim that ordinary desires (e.g. for Bob to be happy) are conditional on their not having terrible consequences (e.g. a billion deaths). But that just seems false to me. I still want Bob to be happy, it's just that I want to avoid a billion deaths even more. The former desire is outweighed, not cancelled. (One might restrict the term 'desire' to non-outweighed or 'all things considered' desires; but I can't imagine why we'd want to. I'm still going to lament Bob's unhappiness; this is importantly different from my cancelled desire for icecream.)
In general, I think it is tempting to over-apply the conditional apparatus. In class, someone suggested the example of a desire for tea on the condition that there's sugar. But this is more plausibly seen as an unconditional desire for tea and sugar. The absence of sugar doesn't cancel the desire, it thwarts it. (Again, the test is to ask whether lacking the object of the desire is in some sense bad from the perspective of the person with the desire.)
One thing I'm unclear on is precisely how a conditional desire (for P given Q) differs from the preference for (Q&P) > (Q&~P). The authors suggest that the conditional desire implies the preference, but not vice versa. Let Q be that you crave chocolate, and P be that you eat chocolate. If you are stuck craving chocolate, it's certainly preferable that you get to have some. You should have the above preference, then. But this needn't imply any actual desire - even conditionally - for chocolate. That sounds right, but then what more is required for desire, exactly? (Not a desire for the condition to obtain, else it might as well be an unconditional desire for P & Q.) Some current pro-attitude towards the object, perhaps? This just seems a little unclear to me.