ARM: Rewrite disassembler to use hsdis

[Andrew Hughes]

----- Original Message -----
> On 01/24/2012 12:43 PM, Andrew Hughes wrote:
> > ----- Original Message -----
> >> On 01/23/2012 09:12 PM, Andrew Hughes wrote:
> >>> ----- Original Message -----
> >>>> On 01/23/2012 02:37 PM, Andrew Hughes wrote:
> >>>>>
> >>>>> I'm very wary about introducing Makefile changes into a release
> >>>>> branch
> >>>>> at such a late stage.  How does these changes affect other
> >>>>> architectures?
> >>>>
> >>>> Not at all.
> >>>
> >>> These chunks would certainly seem to:
> >>>
> >>> +-CFLAGS		+= -fPIC
> >>> ++CFLAGS		+= -fPIC -g
> >>>
> >>> +-CPPFLAGS	+= -I$(BINUTILSDIR)/include -I$(BINUTILS)/bfd
> >>> -I$(TARGET_DIR)/bfd
> >>> ++CPPFLAGS	+= -I$(BINUTILSDIR)/include -I$(BINUTILSDIR)/bfd
> >>> -I$(TARGET_DIR)/bfd
> >>>
> >>> +-	if [ ! -f $@ ]; then cd $(TARGET_DIR); make all-opcodes; fi
> >>> ++	if [ ! -f $@ ]; then cd $(TARGET_DIR); make all-opcodes
> >>> "CFLAGS=$(CFLAGS)"; fi
> >>>
> >>> As far as I can tell from the hard-to-read patch-within-a-patch,
> >>> this part isn't within
> >>> an ARM #ifdef.
> >>
> >> Sure, but hsdis didn't build at all on any arch, so I'm pretty
> >> sure
> >> that other arches aren't affected, unless you count "now builds"
> >> as
> >> an affect.
> >>
> >>> Have you posted this upstream yet?  I'd be interested to hear
> >>> Kelly's thoughts.
> >>>
> >>>>> Is hsdis part of a regular build?
> >>>>
> >>>> No, hsdis is never built in any OpenJDK release.
> >>>
> >>> Ok, so how it is built?  Just via icedtea-debug?
> >>
> >> It's built by the user, if they want disassembly.  It's not part
> >> of any OpenJDK build.  It's not built for some time on IcedTea, as
> >> far as I can tell.
> > 
> > So what I gather from this is that if you want to build hsdis, you
> > have
> > to manually run make on the directory containing the hsdis
> > Makefile. Is
> > that correct?
> 
> Yes.  Also, HSDIS has often needed a bit of local makefile tweaking.

Right.  This is the first I've heard of this tool.

> 
> > If so, I can see that this may break building it in some cases
> > where it
> > worked before, but, from what you imply, the likelihood of this
> > makes the
> > risk negligible.
> > 
> > I do still think the Makefile changes need to go upstream.
> 
> I see that, but HSDIS is quite different in the upstream HotSpot
> trunk,
> so can't go in directly.  Really, the only reason for these changes
> in
> IcedTea is for the convenience of IcedTea developers.
> 
> It's on my list to check out that upstream HotSpot trunk, see what is
> needed for ARM, and submit those changes upstream.
> 

On that subject, what's the plan for getting the ARM port into IcedTea7?
Doing further work in IcedTea6 post-1.11 would seem like a dead end with
distros like Fedora planning to go 1.7 only.

For reference, IcedTea6 is currently using hs20b11, while jdk7u1/IcedTea7 2.0 uses
hs21b17 and jdk7u2/IcedTea7 2.1 will use hs22b10.  So further porting work
may be required.

> Andrew.
> 

-- 
Andrew :)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

PGP Key: 248BDC07 (https://keys.indymedia.org/)
Fingerprint = EC5A 1F5E C0AD 1D15 8F1F  8F91 3B96 A578 248B DC07