[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: java bytecode / java runtime version mismatch



On Mon Nov 10 10:51, Michael Tautschnig wrote:
> > On Mon, Nov 10, 2008 at 3:46 AM, Matthew Johnson <mjj29@debian.org> wrote:
> > <snip>
> > > That version uses javap -verbose, which is rather slow, other
> > > suggestions welcome. It uses the following table:
> > >
> > > highest version in the package => depends line
> > >
> > > 50 => java-runtime6
> > > 49 => java-runtime6 | java-runtime5
> > > 48 => java-runtime6 | java-runtime5 | java-runtime2
> > > <48 => java-runtime6 | java-runtime5 | java-runtime2 | java-runtime
> > </snip>
> > 
> > This is really nice except that you have got the dependency names
> > wrong. They should be java6-runtime, java5-runtime, java2-runtime etc.

Oh, whoops, good catch, I'll rev the version in experimental tonight
with this

> > 
> > And have you considered the possibility of having dependency only on
> > headless runtimes? I am not sure how easy it is to detect as you will
> > need to check what package each class file depends on ex. swing or awt
> > etc.
> > 

Yeah, there's an option for headless in there


> Either way, I think you must also add real packages to your Depends, all the
> above are only virtual ones.

Yeah, I know this, the real package is default-jre, default-jre-headless
or gcj-compat-dev depending on a few things (those have teh above
alternates) or you can specifically depend on openjdk or a sun VM in
which case you don't get the alternates.

Library packages don't keep depends on runtimes (as per policy), only
application packages do. This is to avoid unnecessary deps

Matt

-- 
Matthew Johnson

Attachment: signature.asc
Description: Digital signature


Reply to: