- Yhc - Haskell.org

Added stuff to Prelude
by Dimitry Golubovsky 11 Jan '07

11 Jan '07

Hi, I've done it as Malcolm suggested, that is, manually. Now we have instance Functor Maybe. PS I tested full recompilation of both packages: worked for me. -- Dimitry Golubovsky Anywhere on the Web

1 0

Re: Yhc Digest, Vol 14, Issue 3
by Christopher Lane Hinson 10 Jan '07

10 Jan '07

> An instance Functor Maybe was missing, and I tried to add it to > Prelude (where both Functor and Maybe are defined, to avoid orphan > instance issues). It turns out, every time Prelude.hs is recompiled, > its interface file is overwritten with one from bootstrap. What is the > purpose of this? I'd assume the opposite: if Prelude got recompiled, > the bootstrap interface has to be updated, so one getting a clean > distribution would have prelude interface intact. Is this an error in > SConscript? Or am I missing something? > > For now, I just commented out three lines in SConscript responsible > for the yhc-base package compilation. The new instance is now visible. > > I haven't committed this to the central repo; I'd like to get some > explanation of this first. I also haven't submitted a patch that also had to implement instance Functor Maybe. We seem to have a very similar problem. >> Apparently there are problems with orphan instances, and perhaps >> passing -prelude when compiling the file with the instance in it will >> work. If that doesn't work email the list, since its probably an nhc >> issue originally. For clarity: I did a quick test. I think that orphan instances work when they involve non-prelude names. What I was told by someone else is that all names that appear in the prelude might not be exported unless -prelude, which is a different problem. I originally suggested that orphan instances didn't work because it didn't occur to me that prelude names would be given special treatment. --Lane

1 0

Adding stuff to Prelude (was: Trying to compile Fudgets with Yhc...)
by Dimitry Golubovsky 10 Jan '07

10 Jan '07

Hi, Following up my own message about problems compiling Fudgets with Yhc. An instance Functor Maybe was missing, and I tried to add it to Prelude (where both Functor and Maybe are defined, to avoid orphan instance issues). It turns out, every time Prelude.hs is recompiled, its interface file is overwritten with one from bootstrap. What is the purpose of this? I'd assume the opposite: if Prelude got recompiled, the bootstrap interface has to be updated, so one getting a clean distribution would have prelude interface intact. Is this an error in SConscript? Or am I missing something? For now, I just commented out three lines in SConscript responsible for the yhc-base package compilation. The new instance is now visible. I haven't committed this to the central repo; I'd like to get some explanation of this first. Thanks On 1/10/07, Neil Mitchell <ndmitchell(a)gmail.com> wrote: > Dimitry, > > Apparently there are problems with orphan instances, and perhaps > passing -prelude when compiling the file with the instance in it will > work. If that doesn't work email the list, since its probably an nhc > issue originally. -- Dimitry Golubovsky Anywhere on the Web

1 0

Trying to compile Fudgets with Yhc...
by Dimitry Golubovsky 10 Jan '07

10 Jan '07

Hi, I am trying to get some pieces of Fudgets work in Javascript, and tried to compile some basic modules (stream processor primitives). My goal is to be able to compile unmodified code in its current (pre-cabalized) layout (there are more than 300 modules in Fudgets, so it is easy to break something if changing a lot) I cannot compile EitherUtils.hs http://www.golubovsky.org/repos/Fudgets/hsrc/utils/EitherUtils.hs ========================== Compiling EitherUtils ( ../../../home/dima/ext/workdirs/Fudgets/hsrc/utils/EitherUtils.hs ) yhc: The class Prelude.Functor has no instance for the type Prelude.Maybe. Possible sources for the problem are: 7:13-7:16 When type checking declarations at: 7:1-7:34 ========================== The line in question is: plookup p = fmap snd . find (p.fst) Commenting it out makes the module compilable, but plookup is indeed used somewhere. There was no needed instance in Prelude.hs. I added it (from Hugs, but I believe it is standard): instance Functor Maybe where fmap _ Nothing = Nothing fmap f (Just a) = Just (f a) and then recompiled everything under packages. The problem did not disappear. I even replaced `import Maybe' with `import Data.Maybe', it did not help. I moved the instance definition to Data/Maybe.hs, again no help. What could it be? Thanks. PS Fudgets were compilable by NHC98, weren't they? -- Dimitry Golubovsky Anywhere on the Web

1 0

buildbot failure in linux_ppc
by andrewjwilkinsonw＠gmail.com 09 Jan '07

09 Jan '07

The Buildbot has detected a failed build of linux_ppc. Full details are available at: http://www.indiegigs.co.uk:8010/linux_ppc/builds/29 Buildbot URL: http://www.indiegigs.co.uk:8010/ Buildslave for this Build: linuxppc Build Reason: Build Source Stamp: HEAD Blamelist: Neil Mitchell BUILD FAILED: failed yhi sincerely, -The Buildbot

1 0

Re: [Yhc] Visual Studio 2005 gives libc.lib missing errors
by Peter Tanski 08 Jan '07

08 Jan '07

On Jan 8, 2007, at 1:05 PM, Peter Tanski wrote: > There are also binaries under the ClamAV Native Win32 Port, at > http://oss.netfarm.it/clamav/ I was wrong: those aren't binaries. Here are the directions: 1. download and install MASM if you don't have it yet (it is a small, single executable; tiny install). Download from: http://www.microsoft.com/downloads/details.aspx? FamilyId=7A1C9DA0-0510-44A2-B042-7EF370530C64&displaylang=en note: it installs into the VC++/bin directory 2. download and install 7zip if you don't have it. Download from: http://www.7-zip.org/ 3. download the full GMP library (currently v4.2.1) and extract it into a directory, i.e., \gmp-4.2.1 http://swox.com/gmp/#DOWNLOAD 4. download the appropriate archive from http://oss.netfarm.it/ clamav/ and extract it *into* the gmp directory. 5. the .vcproj file will be in \gmp-4.2.1\msvc\gmp. Open it. 6. in normal VC++ I believe you may be fine; in VC++ Express you need to do this: * go to the project Properties and under Resources/General, * under Additional Include Directories add the include directory that contains "windows.h" (it is included in the resource file). VC+ + Express does not come with "windows.h," but I believe the "windows.h" from Mingw should work. After all this it should build fine as a DLL (gmpd.dll). Note that the default Build is a Debug Build, so you might want to change that as well. Cheers, Pete P.S., there might be an easier way to automate this process (especially finding a "windows.h"). Since it is only a resource file that needs windows.h, it might be possible to simply repackage the whole distribution (gmp, the gmp-msvc version and windows.h) together. It might not introduce conflicts if someone has another windows.h in their path (i.e., full version or old version of VC++) but the local directories are searched first. If all that works you might be able to use scons to call the build.

1 1

Visual Studio 2005 gives libc.lib missing errors
by Peter Tanski 08 Jan '07

08 Jan '07

Neil Mitchell wrote: > I just tried to compile Yhi on Visual Studio 2005. It failed because > it is missing libc.lib, which MS removed for this version. I managed > to fake this by going to VC\lib and copying libcmt.lib and calling it > libc.lib. I suspect the issue is that the copy of gmp we use > explicitly asks for libc.lib, which breaks this. To my knowledge, VS 2005 does not contain the static runtime libraries, although the Platform SDK does contain some static libraries, such as libcp.lib. So a couple of questions: 1) Anyone know a version of gmp which will work with all VS versions? There are precompiled .DLL versions of GMP for Windows and directions for compiling it as a DLL under cygwin and patch files for compiling under VC++.NET at: http://cs.nyu.edu/exact/core/gmp/ Cheers, Pete

2 2

Re: [Yhc] VC(++) environment variables (vcvars32.bat, Issue 49)
by Peter Tanski 08 Jan '07

08 Jan '07

Hi Andrew, Thanks! Andrew Wilkinson wrote: > On 1/7/07, Peter Tanski wrote: >> I looked at Issue 49, on setting the correct environment variables >> for a Visual Studio (Visual C++) toolset using vcvars32.bat or >> whatever the correct batch file is (see below)... > > ... Getting the right environment variables is probably complicated > enough to warrant a new file with a function called from build/ > platform.py, > but it could be included in the same function. It would also be easier to modify later on. Am I correct that "vc" is the MS CL compiler? >> The basic strategy I want to try is this: >> >> 1. find the correct Visual Studio build by querying the system ... > > The dictionary os.environ is probably what you're looking for. > >> 2. find the correct vcvars32.bat or vsvars32.bat ... The easiest >> solution is to query the settings in the Registry... I haven't >> explored this thoroughly and expect there is some cool Python or >> scons function for that purpose.) > > http://sourceforge.net/projects/pywin32/ will let you query the > registry. > I'm not sure if it is included in the main Python windows > installer, but it > is definitely include in ActiveState's ActivePython. The Python 2.5 Windows installer includes _winreg, as well. > 3. run the batch file and query the environment variables > > Is the batch file simple enough to use regular expressions to > extract the > values? That might be easier? It could be parsed but in many cases and for different versions of VS, the main batch file calls other batch files. I discovered a MS tool, Program Files\Microsoft Platform SDK [version, type]\Setup \VCDetect.Exe that will find the correct vsvars32.bat. > 4. propagate the variables into scons ENV > >> ... after this, if Yhc ever needs the Visual Studio tools again, it >> might be able to perform the same process after (or during) >> installation. > > This is as simple as > env["CC"] = "path to compiler" Very nice :) Cheers, Pete

1 0

Visual Studio 2005 gives libc.lib missing errors
by Neil Mitchell 08 Jan '07

08 Jan '07

Hi, I just tried to compile Yhi on Visual Studio 2005. It failed because it is missing libc.lib, which MS removed for this version. I managed to fake this by going to VC\lib and copying libcmt.lib and calling it libc.lib. I suspect the issue is that the copy of gmp we use explicitly asks for libc.lib, which breaks this. So a couple of questions: 1) Anyone know a version of gmp which will work with all VS versions? 2) Has anyone compiled Yhc with VS 2005 before? Thanks Neil

1 0

VC(++) environment variables (vcvars32.bat, Issue 49)
by Peter Tanski 08 Jan '07

08 Jan '07

[redirected from cvs-ghc list] On Jan 7, 2007, at 3:03 PM, Neil Mitchell wrote: > Peter Tanski wrote: >> Ah. Nothing like a little exploration :) Does ./build/platform.py >> sound like the right place to do it? > > That's well beyond my scons knowledge. If you repost the question to > the yhc@ list I'm sure our scons expert will be able to give you an > idea. To carry the discussion we were having, here is the summary: I looked at Issue 49, on setting the correct environment variables for a Visual Studio (Visual C++) toolset using vcvars32.bat or whatever the correct batch file is (see below). This shouldn't be too hard to do but I don't know which SCons files I should modify for this. If any of you have extra time, would you please point me in the right direction? The basic strategy I want to try is this: 1. find the correct Visual Studio build by querying the system and finding the %VSnnCOMNTOOLS% installation. In DOS, this would be as simple as: C:\> set | find "COMNTOOLS" VS80COMNTOOLS=C:\.... 2. find the correct vcvars32.bat or vsvars32.bat (possibly by searching the Start Menu for the Visual Studio [version] Command Prompt) and examining the shortcut. There are only two "Start Menu" systems to check, the current User and "All Users." There is a program, 'getlnk' from Bamboo Software <http://www.inwa.net/~frog/ bamboo.htm> to query the .lnk file in case there isn't a Windows program. (Note: easiest solution is to query the settings in the Registry, under \Software\Microsoft\VisualStudio\[version]\External Tools, I believe. I haven't explored this thoroughly and expect there is some cool Python or scons function for that purpose.) 3. run the batch file and query the environment variables 4. propagate the variables into scons ENV ... after this, if Yhc ever needs the Visual Studio tools again, it might be able to perform the same process after (or during) installation. >> The vcvarsall.bat in the \Visual Studio X\VC\ directory has goto's >> that essentially redirect to the \Visual Studio X\Common7\Tools >> \vsvars32.bat; my vsvars32.bat doesn't have any goto's in it. > > The one in my VC98 installation has some if's in it, so a little more > complex than just set's. It's also named vcvars32.bat instead of > vSvars32.bat. Thanks very much for any any help you can give. Cheers, Pete

2 1