file dialog aborts on amd64 6-RC1

[Sean McNeil]

On Fri, 2005-10-14 at 17:12 -0700, Sean McNeil wrote:
> On Fri, 2005-10-14 at 12:51 -0600, Greg Lewis wrote:
> > On Fri, Oct 14, 2005 at 09:57:29AM -0700, Sean McNeil wrote:
> > > On Fri, 2005-10-14 at 10:50 +0300, Panagiotis Astithas wrote:
> > > > Sean McNeil wrote:
> > > > > I tried to do an import -> plugins and browse the directory location.
> > > > > This consistenly causes the dialog to crash:
> > > > > 
> > > > > #
> > > > > # An unexpected error has been detected by HotSpot Virtual Machine:
> > > > > #
> > > > > #  SIGSEGV (0xb) at pc=0x00000008021e9c1a, pid=24587, tid=0x51a000
> > > > > #
> > > > > # Java VM: Java HotSpot(TM) 64-Bit Server VM
> > > > > (1.5.0-p2-root_13_oct_2005_18_05 mixed mode)
> > > > > # Problematic frame:
> > > > > # C  [libzip.so+0x7c1a]  inflateEnd+0x3a
> > > > > #
> > > > > # An error report file with more information is saved as
> > > > > hs_err_pid24587.log
> > > > > #
> > > > > # If you would like to submit a bug report, please write
> > > > > # a letter to freebsd-java at FreeBSD.org mailing list
> > > > > #
> > > > 
> > > > As the message says, this should go to freebsd-java, along with the 
> > > > hs_err_pid24587.log file. Have you tried increasing the maximum heap size?
> > > 
> > > OK, I've move this thread to java at .  I figured this predated the eclipse
> > > mailing list and the list threw out my attachment of the log for some
> > > reason.  Also, as I've reported before on a different thread, I believe
> > > there is an issue with eclipse using the wrong API for libz.  There are
> > > problems with eclipse and gtk themes that use the pixmap engine.  It
> > > complains of libz versioning.
> > > 
> > > Now, I have a hard failure that is caused by libz and I've noticed
> > > something peculiar.  It appears
> > > that /usr/local/jdk1.5.0/jre/lib/amd64/libzip.so is invoking a routine
> > > in libz where is should not. libzip doesn't depend on libz.  Might there
> > > be some problem with namespace pollution going on here?  Does libzip
> > > need to be removed from the java build and libz used instead?  This
> > > would appear to be the case as they are not compatible yet are both used
> > > because of gtk.
> > 
> > The problem is that libzip.so needs to have its symbols privatised.  A
> > quick fix for this can be found at:
> > 
> > http://docs.freebsd.org/cgi/getmsg.cgi?fetch=46004+0+archive/2005/freebsd-java/20050807.freebsd-java
> > 
> > Your post jogged my memory on this.  I need to look into it again.
> 
> Bingo!!
> 
> Thanks for the pointer.  I replaced the following patch and rebuilt.
> Now I do not get the abort.  There is probably a better way to do this,
> but someone more qualified can figure this out.
> 
> files/patch-j2se::common::Defs-bsd.gmk
> 
> --- ../../j2se/make/common/Defs-bsd.gmk.orig    Fri Oct 14 15:21:48 2005
> +++ ../../j2se/make/common/Defs-bsd.gmk Fri Oct 14 15:25:30 2005
> @@ -180,7 +180,7 @@
>  ifeq ($(TRUE_PLATFORM), NetBSD)
>  DEVTOOLS_PATH = /usr/pkg/bin/
>  else
> -DEVTOOLS_PATH = /usr/local/bin/
> +DEVTOOLS_PATH = ${LOCALBASE}/bin/
>  endif
>  # Allow the user to override the default value...
>  ifdef ALT_DEVTOOLS_PATH
> @@ -294,7 +294,7 @@
>  PIC_CODE_SMALL = -fpic
>  GLOBAL_KPIC = $(PIC_CODE_LARGE)
> 
> -CFLAGS_COMMON   += $(GLOBAL_KPIC) $(GCC_WARNINGS) -pipe
> +CFLAGS_COMMON   += $(GLOBAL_KPIC) $(GCC_WARNINGS) -pipe -DZ_PREFIX
> 
>  # For most platforms, do not omit the frame pointer register usage.
>  #    We need this frame pointer to make it easy to walk the stacks.
> @@ -311,7 +311,7 @@
>  CFLAGS_OPT      = $(POPT)
>  CFLAGS_DBG      = -g
> 
> -CXXFLAGS_COMMON = $(GLOBAL_KPIC) -DCC_NOEX $(GCC_WARNINGS)
> +CXXFLAGS_COMMON = $(GLOBAL_KPIC) -DCC_NOEX $(GCC_WARNINGS) -DZ_PREFIX
>  CXXFLAGS_OPT   = $(POPT)
>  CXXFLAGS_DBG   = -g

P.S.

This fixed my problem with gtk themes looking like crap because of the
libz mismatch.

Cheers,
Sean