[vmkit-commits] [PATCH] Always add bootclasspath, regardless of precompiled state

Nicolas Geoffray nicolas.geoffray at gmail.com
Wed Oct 26 14:23:32 PDT 2011


On Wed, Oct 26, 2011 at 11:13 PM, Will Dietz <wdietz2 at illinois.edu> wrote:

> On Wed, Oct 26, 2011 at 3:02 PM, Nicolas Geoffray
> <nicolas.geoffray at gmail.com> wrote:
> > On Wed, Oct 26, 2011 at 12:41 AM, Will Dietz <wdietz2 at illinois.edu>
> wrote:
> >>
> >> Inlined below.
> >>
> >> This patch does raise the following question:
> >>
> >> When precompiling, is the goal/claim that *every* class offered from
> >> the boot classpath is compiled?
> >
> > No. And yes :) The classes are actually not compiled. All classes are
> just
> > stored in binary in the generated executable. The reason for doing that
> is
> > that you avoid parsing a zip file at startup, so startup is faster.
> >
>
> Heh, that's right.  Just wondering, why *don't* we precompile the code
> (as well as have the bytes available)?
>

Because it would blow up the executable. j3 just precompiles the code that
is needed for bootstrapping, and little other things (System.out). I don't
want to precompile things like corba, which is in the boot classpath.


>
> >>
> >> If not, then this patch makes sense--if so, then I have some work to do
> >> :).
> >
> > I think you have some work to do :) But you should know the boot
> classpath
> > when compiling J3 right?
> >
>
> You mean when building Precompiled.a/Bootstrap.a?  Yes, and maybe that
> already does what's needed.  From what you're saying, this patch is
> wrong and only would mask a bug elsewhere which can be addressed
> separately.
>

No not a bug. It's fine to analyze the boot classpath, it's just  that it
adds up in startup time.


> >> +
> >> +  // Analyze the boot classpath
> >> +  analyseClasspathEnv(bootClasspathEnv);
> >
> > This should really stay in the if (!bootLoaded) test. It actually parses
> the
> > zip files on the boot classpath, which does have an impact on startup
> time.
> >
>
> Great! And this is why I'm glad you're reviewing these patches :D, and
> what you say makes sense so thank you for that :).
>

No problem, happy I can help! I am also looking forward to use OpenJDK with
VMKit!

Cheers,
Nicolas


>
> ~Will
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/vmkit-commits/attachments/20111026/d36dfca5/attachment.html>


More information about the vmkit-commits mailing list