On 21/05/2010, at 00:17, Ian Lynagh wrote: > On Thu, May 20, 2010 at 11:08:45PM +1000, Roman Leshchinskiy wrote: >> On 20/05/2010, at 23:00, Ian Lynagh wrote: >> >>> On Thu, May 20, 2010 at 02:26:08PM +1000, Roman Leshchinskiy wrote: >>>> This happens on both 10.5 and 10.6. It dies with: >>>> >>>> ld: symbol dyld_stub_binding_helper not defined (usually in >>>> crt1.o/dylib1.o/bundle1.o) >>>> collect2: ld returned 1 exit status >>> >>> Someone else in #ghc saw this, but validate works for me. >> >> That's strange. It happens for me on two different machines and Ben gets >> this on his Mac, too. What should I look at? Is validate actually supposed >> to build the RTS as a dynamic library? > > Yes. > > My build log is here: > http://darcs.haskell.org/~igloo/mac-log.txt > and here's the stderr of that command with -v: > http://darcs.haskell.org/~igloo/with-v.txt
It seems I spoke too soon, the machine that I thought was running 10.5 has been upgraded to 10.6. Are you running 10.5? In any case, your gcc (4.0.1) is older than mine (4.2.1). The problem seems to be that ghc passes these flags to gcc: -Wl,-macosx_version_min -Wl,10.5. If I remove them it links ok. In fact, gcc always seems to pass -macosx_version_min 10.6.3 to collect2 in addition to ghc's flags. IIUC, it is supposed to link in /usr/lib/dylib1.o automatically but doesn't with ghc's additional macosx_version_min. I'm not sure why this happens. Roman _______________________________________________ Cvs-ghc mailing list [email protected] http://www.haskell.org/mailman/listinfo/cvs-ghc
