RE: GHC 6.4 release candidates available (breakage on x86-64)

4 Mar 2005


      On 02 March 2005 05:08, Brian Strand wrote:
...
Donald Bruce Stewart wrote:
...
bstrand:
...
Donald Bruce Stewart wrote:
...
bstrand:
...
Simon Marlow wrote:
...
Just to let you know, there are a number of open bug reports for
GHC on the x86_64 platform, which seem to indicate some kind of
occasional memory/GC problem.  I'm probably not going to be able
to track this down until after the 6.4 release, but we'll put
out a patch as soon as we have one.
Here's an error building ghc-6.4.20050224ghc-6.4.20050224 via
Anders' Debian x86-64 ghc, although it doesn't look like a
memory/GC problem to my untrained eye:
==fptools== make boot - --no-print-directory -r;
in /home/bstrand/src/ghc-6.4.20050224/ghc/lib/compat

...
...
...
...
...
../../../glafp-utils/mkdependC/mkdependC -f .depend   
-I../../includes -- -O    -- cbits/directory.c cbits/rawSystem.c
/usr/bin/ghc -M -optdep-f -optdep.depend  -osuf o    -H16m -O
-H64m -O -I. -Rghc-timing -I../../../libraries -fglasgow-exts
-no-recomp Compat/Directory.hs Compat/RawSystem.hs
Distribution/Compat/ReadP.hs Distribution/Extension.hs
Distribution/GetOpt.hs Distribution/InstalledPackageInfo.hs
Distribution/License.hs Distribution/Package.hs
Distribution/ParseUtils.hs Distribution/Setup.hs
Distribution/Version.hs System/Directory/Internals.hs <>
make all
rm -f System/Directory/Internals.o; if [ ! -d
System/Directory/Internals_split ]; then mkdir
System/Directory/Internals_split; else /usr/bin/find
System/Directory/Internals_split -name '*.o' -print | xargs rm -f
__rm_food; fi; /usr/bin/ghc -H16m -O -H64m -O -I. -Rghc-timing 
-I../../../libraries -fglasgow-exts -no-recomp -split-objs    -c
System/Directory/Internals.hs-o System/Directory/Internals.o 
-ohi System/Directory/Internals.hi warning: don't know how to
split object files on this architecture <> ( cd System/Directory/Internals_split; rm
-f ld.script; touch ld.script; echo "INPUT(" *.o ")" >>ld.script;
/usr/bin/ld -r -x -o ../Internals.old.script; );
/usr/bin/ld:ld.script: file format not recognized; treating as
linker script /usr/bin/ld:ld.script:1: syntax error 
make[4]: *** [System/Directory/Internals.o] Error 1
make[3]: *** [boot] Error 2
make[2]: *** [boot] Error 1
make[1]: *** [boot] Error 1
make[1]: Leaving directory
`/home/bstrand/src/ghc-6.4.20050224/ghc' make: *** [build] Error 1
Can you try building with SplitObjs=NO ?
Wolfgang, Ryan - that looks like a splitter problem, no?
(The splitter is more of a dark art than the evil mangler, I
propose we name it the "diabolical splitter" from now on.)
-- Don
Using SplitObjs=NO gives out many warnings then finally dies with:
../../ghc/compiler/ghc-inplace -optc-O -optc-Wall -optc-W
-optc-Wstrict-prototypes -optc-Wmissing-prototypes
-optc-Wmissing-declarations -optc-Winline -optc-Waggregate-return
-optc-Wbad-function-cast -optc-I../includes -optc-I. -optc-Iparallel
-optc-DCOMPILING_RTS -optc-fomit-frame-pointer -H16m -O -H64m -O -O2
-static -I. -#include Prelude.h -#include Rts.h -#include RtsFlags.h
-#include RtsUtils.h -#include StgRun.h -#includeSchedule.h
-#include Printer.h -#include Sanity.h -#include STM.h -#include
Storage.h -#include SchedAPI.h -#include Timer.h -#include
Itimer.h-#include ProfHeap.h -#include LdvProfile.h -#include
Profiling.h -#include Apply.h -fvia-C -dcmm-lint     -c StgCRun.c
-o StgCRun.o 
In file included from ../includes/Stg.h:149,
                from StgCRun.c:67:
../includes/Regs.h:213: warning: call-clobbered register used for
global register variable ../includes/Regs.h:342: warning:
call-clobbered register used for global register variable
/tmp/ghc13908.s: Assembler messages: /tmp/ghc13908.s:20: Error:
suffix or operands invalid for `jmp' make[2]: *** [StgCRun.o] Error
1 
make[1]: *** [all] Error 1
make[1]: Leaving directory `/home/bstrand/src/ghc-6.4.20050224/ghc'
make: *** [build] Error 1
Are you building unregisterised? Those messages about global register
variables (I think) imply that you are not -- and we haven't yet got
x86_64 (terrible name, I much prefer amd64) registerised. Looks like
you might even have found the point where registerisation breaks.
Try building unregisterised:
    GhcUnregisterised=YES
just as the .hc bootstrap compiler was built.
-- Don
Well, I don't know much about how the bootstrap compiler was built (I
just downloaded it from
http://debian-amd64.alioth.debian.org/pure64/pool/unstable/main/amd64/g/
ghc6/),
...
but building with GhcUnregisterised=YES and SplitObjs=NO died with:
==fptools== make all -wr;
  in /home/bstrand/src/ghc-6.4.20050224/libraries/base
------------------------------------------------------------------------
...
../../ghc/compiler/ghc-inplace -H16m -O -H64m -O -fglasgow-exts -cpp
-Iinclude -"#include" HsBase.h -funbox-strict-fields -ignore-package
base -O -Rghc-timing -fgenerics  -fgenerics    -c GHC/Base.lhs -o
GHC/Base.o  -ohi GHC/Base.hi
ghc-6.4.20050224: internal error: getMBlock: mmap: Invalid argument
     Please report this as a bug to glasgow-haskell-bugs@haskell.org,
     or http://www.sourceforge.net/projects/ghc/
make[2]: *** [GHC/Base.o] Error 254
make[1]: *** [all] Error 1
make[1]: Leaving directory
`/home/bstrand/src/ghc-6.4.20050224/libraries' make: *** [build]
Error 1
You need to remove some leftover .hi-boot and .o-boot files in
libraries.  These were left in the source distribution by mistake, they
won't be there for the final release.

Cheers,
	Simon