Neil Mitchell | 1 May 18:55 2007
Picon

Yhc.Core no longer using Data.Binary

Hi,

Upgrading to Data.Binary was fun, and is clearly the right direction
for the future, but its broken at the moment. The authors hope that
with a bit of luck and a google summer of code project they might have
the required build changes done so that we can use it reliably on
various GHC versions, but that's of no comfort to us for the moment.

As a result, I've backed out the patches moving us to Data.Binary, and
have flipped back to our bad version of Binary - which turns out to
produce 30% smaller files and have similar performance...

Apologies to all, but that means all your Yhc.Core files will be
broken once more - please clean and rebuild.

Several of the buildbots are still down, but I know the reason in each
case and will be working to get them all back to life shortly.

Thanks

Neil
andrewjwilkinsonw | 1 May 22:56 2007
Picon

buildbot failure in linux_ppc

The Buildbot has detected a new failure of linux_ppc.
Full details are available at:
 http://www.indiegigs.co.uk:8010/linux_ppc/builds/115

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 yhc

sincerely,
 -The Buildbot
Neil Mitchell | 2 May 15:53 2007
Picon

Yhc.Core proposed change to primitives

Hi,

I'm considering making a change to Yhc.Core that will effect
primitives and how they are handled. I think the change will make
everything simpler for everyone, but since this is an imcompatible
change, I thought I'd check first.

The current treatment of primitives is basically:

CorePrim String -- a call to a primitive

We also have CoreFunc's which are somewhat primitive, but thats a bit
variable throughout.

My proposed treatment is:

data CoreFunc = CoreFunc ...
                       | CorePrim {corePrimName :: String, corePrimArity :: Int}

Then CorePrim calls become standard CoreFun's.

What are the advantages of my proposal? Firstly, we have a list of all
CorePrim's which are used in the program, and reachability etc. works
on these as standard. Secondly, we currently have at least two ways
primitives show up, now we have just one.

Disadvantages? That Yhc --linkcore needs to magically insert a list of
bytecode primitives into the program. This is a disadvantage only for
the person writing --linkcore (i.e. me), but explains the original
reason why it got done this way.
(Continue reading)

andrewjwilkinsonw | 4 May 09:05 2007
Picon

buildbot failure in gentoo_amd64

The Buildbot has detected a new failure of gentoo_amd64.
Full details are available at:
 http://www.indiegigs.co.uk:8010/gentoo_amd64/builds/150

Buildbot URL: http://www.indiegigs.co.uk:8010/

Buildslave for this Build: gentooamd64

Build Reason: 
Build Source Stamp: HEAD
Blamelist: Neil Mitchell

BUILD FAILED: failed configure_2

sincerely,
 -The Buildbot
andrewjwilkinsonw | 4 May 16:33 2007
Picon

buildbot failure in macosx_x86_dual

The Buildbot has detected a new failure of macosx_x86_dual.
Full details are available at:
 http://www.indiegigs.co.uk:8010/macosx_x86_dual/builds/137

Buildbot URL: http://www.indiegigs.co.uk:8010/

Buildslave for this Build: markus2

Build Reason: 
Build Source Stamp: HEAD
Blamelist: Neil Mitchell

BUILD FAILED: failed failed slave lost

sincerely,
 -The Buildbot
Ricky Barefield | 23 May 01:31 2007
Picon

Core

Hi,

 

I’m doing some work with YHC Core and having a problem when compiling some of my code (which imports import Yhc.Core.Binary) when using GHC.  I need the code as an EXE.  It runs fine in ghci.

 

The complaint is

 

Bad interface file: Yhc/Core/Binary.hi

    mismatched interface file versions: expected 6060, found 6

 

I’ve downloaded YHC again and found Binary.hs, this hasn’t really helped because I can’t remember how I made the .hi file in the first place.

 

If it was by scons build then this has stopped working now it’s asking if I have a C compiler.

 

Basically I’m a little stuck and any help regarding this will be much appreciated.

 

Kind regards,

 

Ricky Barefield

_______________________________________________
Yhc mailing list
Yhc <at> haskell.org
http://www.haskell.org/mailman/listinfo/yhc
Neil Mitchell | 23 May 01:52 2007
Picon

Re: Core

Hi Ricky,

> The complaint is
>
> Bad interface file: Yhc/Core/Binary.hi
>     mismatched interface file versions: expected 6060, found 6
>
> If it was by scons build then this has stopped working now it's asking if I
> have a C compiler.

You need to type:

$ scons clean
$ scons build yhc
$ scons build libraries

That way even if you don't have a C compiler you can still compile Yhc
and all the necessary libraries.

Thanks

Neil
Dimitry Golubovsky | 26 May 06:16 2007
Picon

Javascript backend fixed to compile with new Core Types

Hi,

I have fixed the Javascript backend to accomodate the new type layout
in Core (primitives and functions got an unified type).

Changes have been pushed to the repo. No new functionality was added.
At least this demo program:

http://www.golubovsky.org/repos/hibb/EchoCPS.hs

compiles to this web page:

http://www.golubovsky.org/repos/hibb/EchoCPS.html

Those who experiment with the Javascript backend please try compiling
your programs and let me know if anything went broke.

Thanks.

--

-- 
Dimitry Golubovsky

Anywhere on the Web
Neil Mitchell | 27 May 14:54 2007
Picon

GHC 6.4 support

Hi

I've now tracked down the GHC 6.4 issues we were having, which were
down to a combination of:

* module dependencies and C preprocessing in GHC
* fps going from fps to base (wait for 6.8 til it comes out again...)

All our 6.4.2 buildbots now show success, if you do run into issues on
 6.4.2 let me know.

Thanks

Neil
andrewjwilkinsonw | 29 May 09:40 2007
Picon

buildbot failure in gentoo_amd64

The Buildbot has detected a new failure of gentoo_amd64.
Full details are available at:
 http://www.indiegigs.co.uk:8010/gentoo_amd64/builds/171

Buildbot URL: http://www.indiegigs.co.uk:8010/

Buildslave for this Build: gentooamd64

Build Reason: 
Build Source Stamp: HEAD
Blamelist: Neil Mitchell

BUILD FAILED: failed failed slave lost

sincerely,
 -The Buildbot

Gmane