PING 3: [PATCH FOR REVIEW]: Make source/target options explicit for MakeDeps and jvmti

John Coomes John.Coomes at sun.com
Fri Sep 4 22:40:10 PDT 2009


Andrew John Hughes (gnu_andrew at member.fsf.org) wrote:
> 2009/9/4 John Coomes <John.Coomes at sun.com>:
> > Andrew John Hughes (gnu_andrew at member.fsf.org) wrote:
> >>
> >> Hello?
> >>
> >> Can I push this change or not?
> >
> > I've had two high priority things come up.  I'll take a look later
> > today.
> >
> 
> Ok, thanks John.  It's just a bit worrying when things go all silent :)
> 
> To recap, the changeset passed the Sun build you ran for me, and I
> think we resolved all the issues from the comments.  So, as far as I'm
> aware, it should be okay now, fingers crossed.

Unfortunately, your changes break when builds are done the way many
(most?) long-time HotSpot developers do them, which is

	cd <repo>/make/solaris	# or linux or windows
	gmake fastdebug		# on windows it's nmake debug or build.bat ...

That's the traditional way to make HotSpot, has been for most of a
decade.  The new JAVAC_FLAGS aren't passed down; most likely because
defs.make isn't read.

FWIW, the automated build system (JPRT) starts a build like this:

	cd <repo>/make
	gmake fastdebug

This was added a few years ago.  Would be nice to have 'one build to
rule them all' but others have to agree before we can give up the old
way.

Also, in prior email I suggested:

> ..., if you update JAVAC_COMPILE to include BOOTSTRAP_JAVAC_FLAGS (or
> whatever it is), and the SA makefile also includes -source/-target
> options on the command-line, the latter will override the ones in
> BOOTSTRAP_JAVAC_FLAGS (the SA command line may need to include both
> -source and -target).  At least try it, I think it'll be a simpler change.

-John



More information about the hotspot-dev mailing list