October 2000 - Glasgow-haskell-users

RE: C main
by Sigbjorn Finne 13 Oct '00

13 Oct '00

Simon Marlow writes: > > > > Could someone instruct me on (or point me to) the correct ghc > > command line for calling haskell from C? > > ... > > You can use ghc just like you would use gcc to compile up > your C program, > then use ghc again to link the whole program with the Haskell > module(s). The -no-hs-main flag shouldn't be necessary, as > far as I can tell. eg. > > $ ghc -c main.c > $ ghc -c HaskellStuff.hs > $ ghc main.o HaskellStuff.o > That will (sort of) work if you also insist that HaskellStuff.hs implements Main.main. As the name suggests, -no-hs-main takes care of the case when HaskellStuff.hs doesn't. (Well almost, the 4.08 driver contains a bug in setupLinkOpts(); __init_PrelMain should be handled the same way as PrelMain_mainIO_closure. ) In either case, you need to make sure that the RTS is started up, since you're not using the RTS-supplied main(). The HDirect distribution contains an example of this in examples/server/, but notice that the 4.08 RTS' startupHaskell() signature has changed to also take an extra 'module root' argument. However, even with that change, 4.08 compiled code crashes when running (4.05 works fine), which leads me to believe there's something else going on here :-) This is on the same platform that Brett originally reported the problem on. --sigbjorn btw, 4.05 contained a bug in this area which made -no-hs-main only work when linking with -static. 4.08 seems to have inherited it, i.e., in prepareWin32DllLink(), the condition for pushing Main.dll_o and PrelMain.dll_o onto @Link_file should be ($linking_main && !$NoHaskellMain) and not just $linking_main

2 1

RE: How to get BlockedOnDeadMVar exception?
by Simon Marlow 13 Oct '00

13 Oct '00

> On Fri, Oct 13, 2000 at 04:01:00AM -0700, Simon Marlow wrote: > > Sure it's implemented. However, a main thread will never > get this exception > > (at the moment). Try this example in 4.08.1: > > Funny, it works and almost looks the same like my code. Only > {Ee}xception: > I was using a getLine-loop for preventing the main thread > from exiting. > Could this be the reason I don't get the exception? probably: the exception is generated as a result of the thread being discovered unreachable during garbage collection. If there's no garbage collection, eg. because the program is waiting for I/O, then the exception is never raised. Which brings up an interesting subject: should the RTS start a GC after it has been waiting for I/O for a certain length of time? If so, how long? Cheers, Simon

1 0

RE: How to get BlockedOnDeadMVar exception?
by Simon Marlow 13 Oct '00

13 Oct '00

> I didn't succeed in getting (or catching?) this exception through an > obvious example. Is this exception thrown at all or just NYI? > > Is this exception thrown at all or just NYI? Sure it's implemented. However, a main thread will never get this exception (at the moment). Try this example in 4.08.1: module Main where import Exception import Concurrent main = do id <- myThreadId forkIO (catchAllIO (do m <- newEmptyMVar; takeMVar m) (\e -> raiseInThread id e)) catchAllIO (print (sum [1..1000000])) (\e -> print e) Cheers, Simon

2 1

How to get BlockedOnDeadMVar exception?
by Volker Stolz 13 Oct '00

13 Oct '00

I didn't succeed in getting (or catching?) this exception through an obvious example. Is this exception thrown at all or just NYI? Is this exception thrown at all or just NYI? Volker -- Volker Stolz * stolz(a)i2.informatik.rwth-aachen.de * PGP + S/MIME

1 0

RE: C main
by Simon Marlow 12 Oct '00

12 Oct '00

> Greetings - > > Could someone instruct me on (or point me to) the correct ghc > command line > for calling haskell from C? > From the documentation, it seems like "ghc -fglasgow-exts > -no-hs-main main.c > foo.hs" should work, but > that still gives me a multiple definition of "main". > > I'm running ghc 4.08.1 on Windows NT. You can use ghc just like you would use gcc to compile up your C program, then use ghc again to link the whole program with the Haskell module(s). The -no-hs-main flag shouldn't be necessary, as far as I can tell. eg. $ ghc -c main.c $ ghc -c HaskellStuff.hs $ ghc main.o HaskellStuff.o Cheers, Simon

1 0

RE: Porting to powerpc-linux
by Simon Marlow 12 Oct '00

12 Oct '00

> I'm guessing I didn't do the b1 build right. The compiler it > generated > builds, but doesn't even run. Do I need to add the > unregistered options > throughout the build, or just in the library? Just the libraries and the RTS in b1, and the compiler in b2. You can check that things are working right by compiling up hello world against b1 (remember to add the options when compiling hello world too). Cheers, Simon

1 0

Exposing Haskell functions with string arguments in a DLL
by Lescher Christian 12 Oct '00

12 Oct '00

How can I convert string arguments (null-terminated strings) to Haskell's [Char]? I saw there is some CString module in package lang of GHC, but I didn't found any documentation about it (e.g. packStringIO...?). How must a function f :: String -> String for usage in a DLL look like in order to convert the string parameter and result? (From earlier Haskell functions with BSTR arguments, I already know how to build a (static) DLL. I use GHC4.08.1.) Christian Lescher

1 0

C main
by Brett Letner 11 Oct '00

11 Oct '00

Greetings - Could someone instruct me on (or point me to) the correct ghc command line for calling haskell from C? >From the documentation, it seems like "ghc -fglasgow-exts -no-hs-main main.c foo.hs" should work, but that still gives me a multiple definition of "main". I'm running ghc 4.08.1 on Windows NT. Thanks - Brett Letner RBC Incorporated http://www.revbiz.com mailto:bletner@revbiz.com (913) 385-5700 ext.20 Fax (913) 385-5701

1 0

RE: Porting to powerpc-linux
by Simon Marlow 11 Oct '00

11 Oct '00

> I'm considering taking some effort to port GHC to powerpc > linux. The task > seems moderately doable, given that GHC once ran on the powerpc > architecture. Cool! :) > I guess as a good first question: should I start with 4.08.1, > or should I > venture off with the CVS tree. The driver in CVS seems to have been > written in Haskell, and it seems to be needed to build the > runtime. This > sounds like it would preclude building from .hc files. I would start from 4.08.1, for exactly that reason. We haven't tackled the issue of bootstrapping with the Haskell driver. The plan was to use a handwritten Makefile to build an unregisterised driver from .hc sources using plain gcc, but this isn't done yet. Furthermore, we're about to merge the driver with the compiler proper as part of the GHCi reorganisation, so things are going to be broken for a while... > I've got my x86 linux box ready to build some unregistered .hc files. You'll need two build trees. In the first tree, build the compiler as normal, and the libraries/RTS with -unreg -keep-hc-files-too. You should be able to use these .hc files to try out some simple programs on the PPC. In the second tree, bootstrap the compiler using the compiler from the first tree with -unreg -keep-hc-files-too. Take these .hc files to the PPC and bootstrap them there. Cheers, Simon PS. there may be some difficulties with the -unreg flag: it actually specifies way 'u', so you may get .u_hc files instead of .hc files. It might be a better idea to expand -unreg into its component parts: -optc-DNO_REGS -optc-DUSE_MINIINTERPRETER -fno-asm-mangling -funregisterised -fvia-C

2 1

Porting to powerpc-linux
by David Brown 10 Oct '00

10 Oct '00

I'm considering taking some effort to port GHC to powerpc linux. The task seems moderately doable, given that GHC once ran on the powerpc architecture. I guess as a good first question: should I start with 4.08.1, or should I venture off with the CVS tree. The driver in CVS seems to have been written in Haskell, and it seems to be needed to build the runtime. This sounds like it would preclude building from .hc files. I've got my x86 linux box ready to build some unregistered .hc files. Dave (looking forward to debugging RTS) Brown

1 0