January 2004 - Glasgow-haskell-users

RE: linking problems after switching from GHC5 to GHC6
by Simon Marlow 28 Jan '04

28 Jan '04

> I can't link with a library any longer, after switching from ghc5 to > ghc6.2. I've recompiled the library with the new compiler version, so > that's not the problem. [snip] > ghc -o build/test0 build/test0.o -odir build > `src/lib/hsunix-config --local --libs` > build/test0.o(.text+0x25): In function `__stginit_Main_': > : undefined reference to `__stginit_HsUnix_' > build/test0.o(.text+0xec): In function `Main_main_srt': > : undefined reference to `HsUnix_mainwrapper_closure' > build/test0.o(.text+0x151): In function `Main_main_entry': > : undefined reference to `HsUnix_mainwrapper_closure' > collect2: ld returned 1 exit status > make: *** [build/test0] Error 1 > rm build/test0.o Strange. Could you send us the output from -v? Cheers, Simon

2 1

RE: Generics...(ish)
by Simon Peyton-Jones 28 Jan '04

28 Jan '04

Try the new stuff (in GHC6.2) in the Data.Generics library. In particular the class Data, and its related types Constr and DataType http://www.haskell.org/ghc/docs/latest/html/libraries/base/Data.Generics .Basics.html Ralf and I are writing a paper about the extensions you'll find there, which aren't in our original "scrap your boilerplate" paper. Simon | -----Original Message----- | From: glasgow-haskell-users-bounces(a)haskell.org [mailto:glasgow-haskell-users- | bounces(a)haskell.org] On Behalf Of MR K P SCHUPKE | Sent: 21 January 2004 12:35 | To: glasgow-haskell-users(a)haskell.org | Subject: Generics...(ish) | | | I am currently using template haskell to generate code to work on a | record data type... This however is not very pretty, and I wondered if | there was another way to do what I want in GHC. I need to iterate over | the contents of a record for example: | | data Test = Test { a :: Int, b :: String, c :: Float } | | If I derive Typeable, I can compare the data-type, but not decompose it, | I don't know much about generics, and was wondering if I might be able to | do it that way... | | The functionality I need is equivalent to being able to compare to instances, | for example, consider the type above, and the type below: | | data Test2 = Test2 { x :: Int, y :: String, z :: Float, zz :: String } | | i need to construct a function that could output the differenc between the | types, position at a time, so for the above the output would be: | | Test2 contains extra String field. | | or if "y" was to have the type Float, | | Position 2 : Types differ (Test has String, Test2 has Float) | Test2 contains extra String field. | | | I could do this easily if Typeable was a concrete type just by using | case statements... | | Regards, | Keean Schupke. | _______________________________________________ | Glasgow-haskell-users mailing list | Glasgow-haskell-users(a)haskell.org | http://www.haskell.org/mailman/listinfo/glasgow-haskell-users

1 0

RE: "Talking" with the compiler
by Simon Peyton-Jones 28 Jan '04

28 Jan '04

| Yes, the HEP was one attempt to define the interface. There was a | "HEP-lite" implementation at one point, I believe, but it was never | fully adopted. But in effect, GHCi-as-a-library amounts to a HEP expressed as a Haskell library rather than a COM object. So we're getting there. S

1 0

RE: Problem with ghc on Windows ME
by Simon Marlow 27 Jan '04

27 Jan '04

We looked into this issue today (without access to a Win98/ME box though). The conclusion, I think, is that: - rawSystem is wrongly escaping backslashes before executing the command. - NT-derived flavours of Windows are more forgiving than Win95-derived Windows when it comes to multiple backslashes used as directory separators. There's an easy fix (patch attached). If anyone would like to test the patch and let us know if it helps, that would be greatly appreciated. Cheers, Simon

1 0

Test
by shae＠scannedinavian.com 27 Jan '04

27 Jan '04

test

1 0

RE: "Talking" with the compiler
by Simon Marlow 27 Jan '04

27 Jan '04

> > My plan is to have an API where you can request a :load of a module > > source (perhaps omitting the code generation steps for > speed) and then > > request information about the module, by source location > (GHC now has > > completely accurate source location information in its abstract > > datatype; we did this recently in order to support the Visual Studio > > work). The API will most likely be a derivative of the existing > > compilation manager interface: see > > ghc/compiler/compMan/CompManager.lhs. > > What's about the HEP ? > The paper at haskell.org is a little bit old, and I could not find > newer information about the API or it's use. Yes, the HEP was one attempt to define the interface. There was a "HEP-lite" implementation at one point, I believe, but it was never fully adopted. Cheers, Simon

1 0

Problem with ghc on Windows ME
by andrew cooke 27 Jan '04

27 Jan '04

Hi, I'm trying to help someone compile some code of mine and unfortunately they seem to have hit a problem with ghc. Since this is all second hand info I suspect it contains errors and/or isn't useful, but I'm hoping someone recognises the problem as already known and can help out. ghc runs, and preoduces hi files, but doesn't produce a final executable. No warnings are produced. Typical output: C:\halipeto>ghc --make Demo.lhs -o Demo -O -fglasgow-exts Chasing modules from: Demo.lhs (and that's it). I asked him to compile a simple hello world (hw.hs), and a similar thing happens (no "Linking ...") C:\>type hw.hs main = do putStrLn "hello World" C:\>ghc --make hw Chasing modules from: hw Compiling Main ( hw.hs, hw.o ) C:\>a.out Bad command or file name This is ghc6.2 (downloaded in the last day or two): C:\>ghc --version The Glorious Glasgow Haskell Compilation System, version 6.2 Any suggestions? Thanks, Andrew -- personal web site: http://www.acooke.org/andrew personal mail list: http://www.acooke.org/andrew/compute.html

3 12

RE: "Talking" with the compiler
by Simon Marlow 26 Jan '04

26 Jan '04

> On Mon, 2004-01-19 at 11:34, Simon Marlow wrote: > > For the Visual Studio plugin we're going to need to talk to > GHCi. We > > plan to do this by designing an appropriate API for GHCi > and calling it > > directly; you *could* do it by talking over a pipe, but > it's going to be > > a lot of work (and slow). If you want to do this, please talk to us > > about what API you'd like to see, and we can hopefully implement > > something that will be generally useful. > > I wanted something like that for a Haskell IDE I was working on (not > much progress on it at the moment, but I may pick it up again). > > The main things I wanted was enough information to be able to > implement > "jump to definition". Where you select a symbol in your > editor and move > to where that variable/function/type/class was defined, in the same > module or another module. It would also be useful to find out the > module and package a symbol comes from so that an IDE could > have a good > stab at finding some documentation. > > For that, you'd want an API for wandering through the useful > information > in .hi files. An API corresponding to hugs/ghci's :info > <name>, :browse > <modname>, :type <name> would be a good start. You'd want to > be able to > specify which root module to load up the symbols for, optionally > specifing a search path and expect it to also load up the .hi > files for > any imported modules. This is all stuff that we need for Visual Studio too. VS will typecheck your program as you type, so you'll get errors underlined in the source code. A side effect of this is that it will collect all the information reuqired to implement "jump to definition" and "tell me the type of this identifier" in the editor. My plan is to have an API where you can request a :load of a module source (perhaps omitting the code generation steps for speed) and then request information about the module, by source location (GHC now has completely accurate source location information in its abstract datatype; we did this recently in order to support the Visual Studio work). The API will most likely be a derivative of the existing compilation manager interface: see ghc/compiler/compMan/CompManager.lhs. Cheers, Simon

2 2

RE: Concurrent Haskell on Win32?
by Simon Marlow 26 Jan '04

26 Jan '04

> I am confused about the state of Concurrent Haskell on Win32. > I remember > it didn't work in the past because of I/O blocking issues. IIRC, I > actually verified this by compiling a Haskell web server on > both Linux > and Windows. > > Has this been fixed in the meantime? If not, it would be great to > mention this in the GHC release notes in order to prevent people from > figuring this out the hard way. > > Fixing concurrency would expand Haskell's potential range of > applications tremendously... I'd really love to see that happen. Non-blocking I/O support for Windows was added in 6.0.1. Cheers, Simon

1 0

Re: GHC Source Build documentation BUG
by dons＠cse.unsw.edu.au 26 Jan '04

26 Jan '04

arthurvl: > Hello, > > while trying to adapt the local bootstrap-from-scratch script that allows > us to install a somewhat customized GHC onto our systems starting from > nothing but an installed GCC 2.95, I noticed a bug in the documentation > accompanying the GHC 6.2 release. On > http://www.haskell.org/ghc/download_ghc_62.html there is a note stating > The source distribution needs an installed GHC 4.08.X, 5.X or 6.X to build. > > As 4.08.2 is relatively easily bootstrapped using a C compiler, that is > the base we start off, and thus I figured I could just jump from 4.08.2 > to 6.2. Unfortunately, the compilation itself then states: > :94: #error GHC >= 5.00 is required for bootstrapping GHC > > Annoying, that. Especially given that I *checked* if I could jump from > 4.08.2 to 6.2. > > Doei, Arthur. (One extra stage in the bootstrapping process...) Hmm. SimonM may have something to say about this. Since 6.0.1 you can bootstrap directly `quite easily', as long as you can generate C source, or someone else has done it for you (this is how the OpenBSD ports tree version of GHC builds). So rather than bootstrap from 4.08.2 (which won't even build on some platforms, i.e. sparc/openbsd, netbsd, ia64, mips64), it might be more profitable to start directly with the version you want, 6.2, and generate your own C using the instructions in the docs. Of course, you need to install a binary of GHC first, but once you've got the C source, you can wipe GHC, and build using gcc only. 4.08.2 should die now. -- Don

1 0