Cool article explaining the shift from feathers evolving for flight to feathers evolving for warmth, and flight evolving from there.

http://ngm.nationalgeographic.com/2011/02/feathers/zimmer-te...

The long, hollow filaments on theropods posed a puzzle. If they were early feathers, how had they evolved from flat scales? Fortunately, there are theropods with threadlike feathers alive today: baby birds. All the feathers on a developing chick begin as bristles rising up from its skin; only later do they split open into more complex shapes. In the bird embryo these bristles erupt from tiny patches of skin cells called placodes. A ring of fast-growing cells on the top of the placode builds a cylindrical wall that becomes a bristle.

Reptiles have placodes too. But in a reptile embryo each placode switches on genes that cause only the skin cells on the back edge of the placode to grow, eventually forming scales. In the late 1990s Richard Prum of Yale University and Alan Brush of the University of Connecticut developed the idea that the transition from scales to feathers might have depended on a simple switch in the wiring of the genetic commands inside placodes, causing their cells to grow vertically through the skin rather than horizontally. In other words, feathers were not merely a variation on a theme: They were using the same genetic instruments to play a whole new kind of music. Once the first filaments had evolved, only minor modifications would have been required to produce increasingly elaborate feathers.
--------------------

Until recently it was thought that feathers first appeared in an early member of the lineage of theropods that leads to birds. In 2009, however, Chinese scientists announced the discovery of a bristly-backed creature, Tianyulong, on the ornithischian branch of the dinosaur family tree—about as distant a relative of theropods as a dinosaur can be. This raised the astonishing possibility that the ancestor of all dinosaurs had hairlike feathers and that some species lost them later in evolution. The origin of feathers could be pushed back further still if the "fuzz" found on some pterosaurs is confirmed to be feathers, since these flying reptiles share an even older ancestor with dinosaurs.