
This is to announce that testing for 6.4.1 has begun. Snapshots from 20050730 onwards are release candidates - please test if you can. Snapshots are available from here: http://www.haskell.org/ghc/dist/stable/dist/ we have source snapshots, and binaries for Linux/x86 (RH 9.1 era), Linux/x86_64 (FC3 era), and Windows. A *huge* number of fixes have gone in since 6.4. Also, big news is that Linux/x86_64 will have a new native code generator and GHCi support (including Template Haskell), so we'd especially like to hear from anyone that can test on Linux/x86_64. The GHCi support isn't complete: the interpreter can't currently handle FFI, and there are problems with data references to shared libraries from dynamically-loaded .o files, but otherwise it should be functional. We're planning for a release in 1-2 weeks. I'm going to be offline for a while from mid-Aug, but hopefully we'll get the release out before then. Please remember to put the full version number in any bug report, and happy testing! Cheers, Simon

Simon Marlow wrote:
This is to announce that testing for 6.4.1 has begun. Snapshots from 20050730 onwards are release candidates - please test if you can.
I built ghc-6.4.1.20050731 successfully on Fedora Core 4 on x86_64 with ghc-6.4. I haven't tested it much yet, but ghci works. :-) (Please ask if you're interested in the rpms.) Jens

Simon Marlow wrote:
Linux/x86_64 will have a new native code generator and GHCi support
Great news!
However building 20050801 gives me the following error:
,----[ make ]
| ------------------------------------------------------------------------
| ==fptools== make all -wr;
| in /home/gebner/build/ghc-6.4.1.20050801/libraries/base
| ------------------------------------------------------------------------
| rm -f System/Posix/Signals.o; if [ ! -d System/Posix/Signals_split ]; then
mkdir System/Posix/Signals_split; else /usr/bin/find
System/Posix/Signals_split -name '*.o' -print | xargs rm -f __rm_food; fi;
| ../../ghc/compiler/ghc-inplace -H16m -O -fglasgow-exts -cpp -Iinclude
-"#include" HsBase.h -funbox-strict-fields -ignore-package base -O
-Rghc-timing -fgenerics -fgenerics -split-objs -c
System/Posix/Signals.hs -o System/Posix/Signals.o -ohi
System/Posix/Signals.hi
| In file included from System/Posix/Signals.hs:108:
| /home/gebner/build/ghc-6.4.1.20050801/ghc/includes/Signals.h:18: warning:
extra tokens at end of #endif directive
|
| System/Posix/Signals.hs:395:30: Not in scope: `sizeof_sigset_t'
|
| System/Posix/Signals.hs:401:30: Not in scope: `sizeof_sigset_t'
|
| System/Posix/Signals.hs:408:31: Not in scope: `sizeof_sigset_t'
|
| System/Posix/Signals.hs:411:22: Not in scope: `sizeof_sigset_t'
|
| System/Posix/Signals.hs:417:31: Not in scope: `sizeof_sigset_t'
|
| System/Posix/Signals.hs:420:22: Not in scope: `sizeof_sigset_t'
|
| System/Posix/Signals.hs:432:30: Not in scope: `sizeof_sigset_t'
|
| System/Posix/Signals.hs:453:30: Not in scope: `sizeof_sigset_t'
| <

Gabriel Ebner (ge@gabrielebner.at) wrote:
This is on gentoo/amd64 with gcc-3.4.4 and glibc-2.3.5.
I just built The Glorious Glasgow Haskell Compilation System, version 6.4.1.20050801 on Gentoo/amd64 by bumping version of ebuild to 6.4.1 with gcc-3.4.4 & glibc-2.3.5 and everything is fine...built darcs, gtk2hs & there is ghci :-)) Congratulations to devs!! Sincerely, Gour -- Registered Linux User | #278493 GPG Public Key | 8C44EDCD

Gour wrote:
Gabriel Ebner (ge@gabrielebner.at) wrote:
This is on gentoo/amd64 with gcc-3.4.4 and glibc-2.3.5.
I just built The Glorious Glasgow Haskell Compilation System, version 6.4.1.20050801 on Gentoo/amd64 by bumping version of ebuild to 6.4.1 with gcc-3.4.4 & glibc-2.3.5 and everything is fine...built darcs, gtk2hs & there is ghci :-))
I just tried on debian/amd64 unstable and it worked too. And that was with gcc-4.0! The GHC devs don't cease to impress me. I don't notice any particular tweaking in the ghc ebuild. Did you uncomment the GhcWithInterpreter and SplitObjs lines in the ebuild? I'll try to rebuild ghc on my gentoo box, maybe it works this time... Gabriel.

Gabriel Ebner (ge@gabrielebner.at) wrote:
I just tried on debian/amd64 unstable and it worked too. And that was with gcc-4.0! The GHC devs don't cease to impress me.
:-)
I don't notice any particular tweaking in the ghc ebuild. Did you uncomment the GhcWithInterpreter and SplitObjs lines in the ebuild?
I just tweaked the 2nd line to: use ppc64 && echo "GhcWithInterpreter=NO" >> mk/build.mk so I can get 'ghci' built.
I'll try to rebuild ghc on my gentoo box, maybe it works this time...
Pls. report back if everything is fine. Sincerely, Gour

Gour wrote:
I don't notice any particular tweaking in the ghc ebuild. Did you uncomment the GhcWithInterpreter and SplitObjs lines in the ebuild?
I just tweaked the 2nd line to:
use ppc64 && echo "GhcWithInterpreter=NO" >> mk/build.mk
so I can get 'ghci' built.
I think you can do away with the first line too, I've just done a successful build with split objects on gentoo/amd64.
I'll try to rebuild ghc on my gentoo box, maybe it works this time...
Pls. report back if everything is fine.
I must've touched the magic switch. Now it's running just fine. Apparently HTYPE_TCFLAGS_T got undef'd the first time, but now it's good, although I don't know what I did. Gabriel.

Gabriel Ebner (ge@gabrielebner.at) wrote:
I think you can do away with the first line too, I've just done a successful build with split objects on gentoo/amd64.
Is there any advantage in split-build?
I must've touched the magic switch. Now it's running just fine. Apparently HTYPE_TCFLAGS_T got undef'd the first time, but now it's good, although I don't know what I did.
Glad that everything is fine. Finally x86_64 is becoming a first-class citizen ;) Sincerely, Gour -- Registered Linux User | #278493 GPG Public Key | 8C44EDCD

I just tweaked the 2nd line to:
use ppc64 && echo "GhcWithInterpreter=NO" >> mk/build.mk
so I can get 'ghci' built.
I think you can do away with the first line too, I've just done a successful build with split objects on gentoo/amd64.
I can confirm that GHC 6.4.1.20050801 works on my Gentoo AMD64 box under GCC 4.0.1, with the stock 6.4 ebuild modified only for the version bump and to build with split objects and ghci, as above. Thanks for the good work! :-) I've been looking forward eagerly to native AMD64 support; sure, I've got a 32-bit chroot, but it's just not the same :-D Luke Maurer maurerl@carleton.edu

On Mon, Aug 01, 2005 at 03:07:16PM +0100, Simon Marlow wrote:
This is to announce that testing for 6.4.1 has begun. Snapshots from 20050730 onwards are release candidates - please test if you can. Snapshots are available from here:
http://www.haskell.org/ghc/dist/stable/dist/
we have source snapshots, and binaries for Linux/x86 (RH 9.1 era), Linux/x86_64 (FC3 era), and Windows.
20050801 looks OK so far here on x86 Debian Linux, apart from the building.xml issue already mentioned. debs for i386 unstable are in Haskell Unsafe: http://haskell-unsafe.alioth.debian.org/haskell-unsafe.html Thanks Ian

Hi, ghc-6.4.1.20050804 builds successfully on Mac OS X 10.4.2 using gcc 3.3. This snapshot also appears to fix bug 1249527, "runhaskell causes GHC to segfault on OS X". The new compiler was also tested by building darcs-1.0.3. The build was done under the darwinports infrastructure using ghc-6.4 as the bootstrap compiler. I haven't yet tried to build using gcc 4.0; should I expect that to work yet? Best Wishes, Greg
participants (7)
-
Gabriel Ebner
-
Gour
-
Gregory Wright
-
Ian Lynagh
-
Jens Petersen
-
Luke Maurer
-
Simon Marlow