Johan Tibell wrote:
Why don't we use the DWARF information instead? It has no runtime overhead so it can actually be turned on always. It also integrates with all the standard open source tooling.
I think there are two backtrace-like chains of interest for every Haskell value: There is one "static" chain that tracks how values are created, and one "dynamic" chain that tracks the actual evaluation. For example, for makePair :: (Int, Int) makePair = (error "The first component is actually undefined", 42) sumPair :: (Int, Int) -> Int sumPair (a, b) = a + b main = print $ sumPair makePair the static chain up to the error call would be main -> makePair -> error while the dynamic chain that actually produces the error is something like main -> print -> show -> sumPair -> (+) -> error As far as I understand, DWARF information will give us the dynamic chain, while the ImplicitLocations idea will give us fragments of the static chain. So I believe these two features are complementary, and most useful if combined. Cheers, Bertram