First of all, (StateT Diffractometer (Pipe Engine Geometry IO) ()) isn't the same as (Pipe Engine Geometry (StateT Diffractometer IO) ()), although I'm not sure what exactly the difference will be, as I've never used it the former way. This might be again a question for the Pipes mailing list.
ok, I will investigate this part :)
but now if I remove the get and put lines, I get the segfault. Okay, I have no idea. As I see it, this shouldn't happen, as you're getting and putting the same pointer all the time. What if you remove
Secondly: the StateT altogether and just use the `dif` from the function argument, are you still getting segfaults? Also what about writting the function without using pipes, and using Pipes.mapM to make it a pipe like I mentioned? (if the only Pipes operation you're doing are an `await` in the beginning and a `yield` at the end, Pipes.mapM covers it)
Yes I get the segfault if I remove all the State Part and use directly the dif. I think that I have something like this , I need to connect the life of one foreignPtr to another one. This is why I put them all into the Diffractometer data. But it seems that this is not the magic bullet in order to garanty the same life time for all of these foreignPtr. I am wonderig if the fact that I use only one of the ForeignPtr in the solveTrajPipe function does not give a clu to haskell that he just need to keep a reference to the used one and get ride of the other one stored in dif. withDiffractometer :: Diffractometer -> (Ptr HklEngineList -> IO b) -> IO b withDiffractometer d fun = do let f_engines = difEngineList d withForeignPtr f_engines fun then I extract a pointer from the engines one but as I am using the withForeingPtr I think that it is ok. I would say that I am not yet a specialist of FFI and haskell, but I love this language a lot when I will manage this problem :)) solveTrajPipe' :: Diffractometer -> Pipe Engine Geometry IO () solveTrajPipe' dif = flip evalStateT dif $ forever $ do -- Inside here we are using `StateT Int (Consumer a IO) r` e <- lift await -- dif <- get let name = engineName e solutions <- lift . lift $ withDiffractometer dif $ \engines -> withCString name $ \cname -> do engine <- c_hkl_engine_list_engine_get_by_name engines cname nullPtr n <- c_hkl_engine_pseudo_axis_names_get engine >>= darrayStringLen solve' engine n e >>= getSolution0 -- put dif lift $ yield solutions Cheers Frederic Ps: just for fun I got this adding just a lift print :) :~/hkl/contrib/haskell$ cabal run Preprocessing executable 'ghkl' for hkl-0.1.0.0... [3 of 5] Compiling Hkl.C ( dist/build/ghkl/ghkl-tmp/Hkl/C.hs, dist/build/ghkl/ghkl-tmp/Hkl/C.p_o ) src/Hkl/C.hsc:125:3: Couldn't match kind `* -> *' with `*' Expected type: Diffractometer -> Proxy () Engine () Geometry IO () Actual type: Diffractometer -> Proxy () Engine () Geometry IO () Kind incompatibility when matching types: Diffractometer :: * -> * Diffractometer :: * The function `lift'ghc: panic! (the 'impossible' happened) (GHC version 7.6.3 for x86_64-unknown-linux): kindFunResult ghc-prim:GHC.Prim.*{(w) tc 34d} Please report this as a GHC bug: http://www.haskell.org/ghc/reportabug