June 2002 - Cvs-hugs - Haskell.org

cvs commit: hugs98/src HsFFI.h Makefile.in builtin.c connect.h ffi.c input.c iomonad.c machdep.c parser.y plugin.c prelude.h static.c storage.c storage.h type.c
by Alastair Reid 14 Jun '02

14 Jun '02

reid 2002/06/14 07:41:15 PDT Modified files: src Makefile.in builtin.c connect.h ffi.c input.c iomonad.c machdep.c parser.y plugin.c prelude.h static.c storage.c storage.h type.c Added files: src HsFFI.h Log: Last of the ffi commits. Hopefully won't break anything... Here's how to test it on your machine: make && ./hugs -P../lib:../lib/exts: +G test.hs < /dev/null && cc -I. -shared ../lib/exts/MarshalAlloc.c -o ../lib/exts/MarshalAlloc.so && cc -I. -shared ../lib/exts/MarshalUtils.c -o ../lib/exts/MarshalUtils.so && cc -I. -shared ../lib/exts/Storable.c ../lib/exts/Storable_aux.c -o ../lib/exts/Storable.so && cc -I. -shared test.c -o test.so && ./hugs -P../lib:../lib/exts -H test.hs; stty sane after creating hugs98/src/test.hs containing: import Foreign import Exception import Prelude hiding (read) tests = do putStrLn "\nTesting sin==mysin (should return lots of Trues)" print testSin putStrLn "\nTesting errno" err <- peek errno putStrLn $ "errno == " ++ show err putStrLn "\nTesting puts (and withString)" withString0 "Test successful" puts putStrLn "\nTesting peekArray0" s <- withString0 "Test successful" (peekArray0 '\0') putStr s putStrLn "\nTesting open, read and close" s <- testRead "test.hs" 200 putStrLn s putStrLn "\nTesting open, write and close" testWrite "/tmp/test_write" "Test successful" putStrLn "\nTesting exit" exit 3 foreign import ccall "sin" mysin :: Double -> Double testSin = [ (sin x == mysin x) | x <- [0,0.01 .. 1] ] foreign import ccall safe "static stdlib.h &errno" errno :: Ptr Int withString s = bracket (newArray s) free withString0 s = bracket (newArray0 '\0' s) free withBuffer sz m = do b <- mallocArray sz sz' <- m b s <- peekArray sz' b free b return s foreign import ccall puts :: Ptr Char -> IO Int foreign import ccall "open" open' :: Ptr Char -> Int -> IO Int foreign import ccall "open" open2' :: Ptr Char -> Int -> Int -> IO Int foreign import ccall "creat" creat' :: Ptr Char -> Int -> IO Int foreign import ccall close :: Int -> IO Int foreign import ccall "read" read' :: Int -> Ptr Char -> Int -> IO Int foreign import ccall "write" write' :: Int -> Ptr Char -> Int -> IO Int creat s m = withString0 s $ \s' -> unix "creat" $ creat' s' m open s m = withString0 s $ \s' -> unix "open" $ open' s' m open2 s m n = withString0 s $ \s' -> unix "open2" $ open2' s' m n write fd s = withString0 s $ \s' -> unix "write" $ write' fd s' (length s) read fd sz = withBuffer sz $ \s' -> unix "read" $ read' fd s' sz unix s m = do x <- m if x < 0 then do err <- peek errno ioError $ userError $ s ++ ": " ++ show (x,err) else return x testRead fn sz = bracket (open fn 0) close (flip read sz) testWrite fn s = bracket (open2 fn (512+64+1) 511) close (flip write s) foreign import ccall exit :: Int -> IO () -- Various bits of rubbish. -- foreign import ccall "static stdlib.h exit" (***) :: Ptr Char -> Ptr Char -> IO Int -- -- foreign import ccall safe "static stdlib.h printf" (+++) :: Ptr Char -> Ptr Char -> IO Int -- foreign import ccall safe "static stdlib.h &errno" illegal_foo :: Ptr Int -- -- foreign import ccall safe "wrapper" illegal_bar :: Char -> IO (FunPtr Char) -- foreign import ccall safe "dynamic" illegal_baz :: FunPtr Char -> Char -- foreign export ccall "id_charstar" id :: Ptr Char -> Ptr Char Here's the changes: HsFFI.h Makefile.in HsFFI.h defines the ffi-mandated types. It really ought to be autoconfed (current settings suits an x86/gcc). Makefile updated to include HsFFI.h machdep.c prelude.h Int64 support static.c type.c parser.y input.c storage.c storage.h connect.h New syntax and related frontend changes. ffi.c Tweak code generator to match changes. Extend with many new ffi-supported types builtin.c iomonad.c plugin.c Runtime support. Big change is addition of HugsAPI4 and associated operations. This extends HugsAPI3 (greencard support) in the usual way. Note: I'd like to eliminate all direct greencard support and do a big cleanup ASAP. There's a lot of accumulated cruft in there. Revision Changes Path 1.13 +16 -16 hugs98/src/Makefile.in 1.24 +186 -37 hugs98/src/builtin.c 1.36 +17 -10 hugs98/src/connect.h 1.6 +281 -177 hugs98/src/ffi.c 1.43 +15 -39 hugs98/src/input.c 1.27 +15 -13 hugs98/src/iomonad.c 1.51 +47 -2 hugs98/src/machdep.c 1.33 +14 -44 hugs98/src/parser.y 1.5 +13 -36 hugs98/src/plugin.c 1.35 +4 -2 hugs98/src/prelude.h 1.70 +283 -225 hugs98/src/static.c 1.41 +16 -4 hugs98/src/storage.c 1.37 +26 -17 hugs98/src/storage.h 1.48 +33 -2 hugs98/src/type.c

1 0

cvs commit: hugs98/lib/exts Storable_aux.c
by Alastair Reid 14 Jun '02

14 Jun '02

reid 2002/06/14 07:25:48 PDT Added files: lib/exts Storable_aux.c Log: Support code for Storable.hs - defines teh read/writeFooOffPtr functions. In writing this code, some attempt was made to make it easier to cut and paste definitions so that it would be easier to generate the missing definitions. It'd probably be easier to use the more obvious technique of using cpp to do the job.

1 0

cvs commit: hugs98/lib/exts Foreign.hs
by Alastair Reid 14 Jun '02

14 Jun '02

reid 2002/06/14 07:22:27 PDT Modified files: lib/exts Foreign.hs Log: Updated to pull in the new ffi libraries. Warning: anyone using this new one will be forced to use the new ffi because it pulls in files which contain foreign imports. Revision Changes Path 1.4 +23 -20 hugs98/lib/exts/Foreign.hs

1 0

cvs commit: hugs98/lib/exts Word.hs
by Alastair Reid 14 Jun '02

14 Jun '02

reid 2002/06/14 07:20:30 PDT Modified files: lib/exts Word.hs Log: Bugfix: Word64 has been defined in the Prelude for over a week now so I'm removing the definition from here. Revision Changes Path 1.7 +11 -11 hugs98/lib/exts/Word.hs

1 0

cvs commit: hugs98/lib/exts CTypes.hs FunPtr.hs MarshalAlloc.hs MarshalArray.hs MarshalError.hs MarshalUtils.hs Ptr.hs StablePtr.hs Storable.hs
by Alastair Reid 14 Jun '02

14 Jun '02

reid 2002/06/14 07:19:29 PDT Added files: lib/exts CTypes.hs FunPtr.hs MarshalAlloc.hs MarshalArray.hs MarshalError.hs MarshalUtils.hs Ptr.hs StablePtr.hs Storable.hs Log: First of a series of commits of the new ffi code. Existing uses of GreenCard-generated code should still work (and seems to do so). These are enough of the libraries to run some tests. Some of them (notably CTypes.hs and Storable.hs) are very incomplete _and_ should (but aren't) be configured with autoconf and should (but isn't) be built from the sources in the hierarchial library archive. I'd love it if someone were to tackle either of these problems before I get to them.

1 0

cvs commit: hugs98/src hugs.c
by Sigbjorn Finne 13 Jun '02

13 Jun '02

sof 2002/06/13 15:56:43 PDT Modified files: src hugs.c Log: Flip default for 'T' toggle (and give it a better doc. string + name.) Revision Changes Path 1.84 +5 -5 hugs98/src/hugs.c

1 0

cvs commit: hugs98/src static.c
by Sigbjorn Finne 13 Jun '02

13 Jun '02

sof 2002/06/13 15:01:31 PDT Modified files: src static.c Log: Handle reporting of ambiguous type constructor + class occurrences. Revision Changes Path 1.69 +40 -25 hugs98/src/static.c

1 0

cvs commit: hugs98/src machdep.c
by Alastair Reid 12 Jun '02

12 Jun '02

reid 2002/06/12 11:55:25 PDT Modified files: src machdep.c Log: Fixed dynamic loading implementation for Darwin. Tweaked the code to use a fresh namespace for each loaded module instead of the global namespace. This means that when you lookup _initModule you get a different answer each time. What was happening before was you always got the first _initModule loaded into the namespace which was always the one from StdDIS.so so we kept initializing StdDIS instead of initializing the module we'd just loaded. Someone who knows Darwin better should check this code over. My source of documentation was some random web page instead of a proper manual. My testcase was an X program being run on a machine on which I couldn't run X - so all I know is that it loads successfully but I can't say for sure that it runs ok. Revision Changes Path 1.50 +8 -7 hugs98/src/machdep.c

1 0

cvs commit: hugs98/tests Readme testScript.in
by Alastair Reid 12 Jun '02

12 Jun '02

reid 2002/06/12 09:55:11 PDT Modified files: tests Readme testScript.in Log: Yesterday's fixes to the testScript were scrogged because.... I forgot that you have to commit testScript not testScript.in While I was in there I fixed the header on testScript.in which should read: CONTEXT_DIFF='@CONTEXT_DIFF@' export CONTEXT_DIFF DEV_NULL='@DEV_NULL@' export DEV_NULL HUGSFLAGS=-P../lib:../lib/hugs:../lib/exts export HUGSFLAGS HUGS=../src/hugs Hmmm, maybe the right thing to do is for configure to make a little config script containing nothing but the above and for testScript to source that file. This'd avoid screwups like yesterday. Revision Changes Path 1.5 +3 -0 hugs98/tests/Readme 1.20 +88 -88 hugs98/tests/testScript.in

1 0

cvs commit: hugs98/tests/libs except1.output hugs98/tests/rts ioerror1.output hugs98/tests/tcheck DictHW.output TyErr.output
by Alastair Reid 11 Jun '02

11 Jun '02

reid 2002/06/11 10:55:56 PDT Modified files: tests/libs except1.output tests/rts ioerror1.output tests/tcheck DictHW.output TyErr.output Log: Fixed test suite to match current behaviour. Two changes seem to have happened: 1) The +q flag makes things even quieter. I didn't really look at what was different but it seems like it doesn't report which files it is loading anymore - so the number of lines to ignore got much smaller. 2) When printing IOErrors, an extra carriage return is added to the error message. Revision Changes Path 1.4 +3 -0 hugs98/tests/libs/except1.output 1.3 +2 -0 hugs98/tests/rts/ioerror1.output 1.5 +2 -4 hugs98/tests/tcheck/DictHW.output 1.3 +0 -1 hugs98/tests/tcheck/TyErr.output

1 0