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

Re: RFS: commons-math



Hi Matt,

On Friday 20 March 2009 00:57:35 Matthew Johnson wrote:
> On Thu Mar 19 21:01, Damien Raude-Morvan wrote:
> > > Ok, can I suggest a compromise. Build-depend on default-jdk and then at
> > > build time compile with default-jdk and, if openjdk is installed, use
> > > it to run  the test suite.
> >
> > It would make the build nondeterministic between different host with
> > different installed packages [1]. I'm reluctant to make this change (aka
> > you have to convice me :P.
>
> The _build_ should be the same everywhere (always default-jdk, you said
> that the bytecode was identical, right), but the test suite would only
> run when you had openjdk to stop it taking such a long time).
>
> Hence, the library can still be used with default-jre, it just might be
> slow.
>
> I don't mind too much though. An alternative could be to compile with
> -target 1.5 to ensure you use a classfile version compatible with gcj.

I finally see our misundertanding : I already build all class files with -source 
1.3 and -target 1.3 (class Format 47) as recommanded by upstream. They will 
run fine on any JVM >= 1.3.

Could you upload my package as is ?

-- 
Damien Raude-Morvan / www.drazzib.com

Attachment: signature.asc
Description: This is a digitally signed message part.


Reply to: