Thanks Rémi. Devs,
I'll merge our change in this area (just the ASM_API_VERSION) from Groovy 3 into the GROOVY_2_5_X branch unless there are any objections. Cheers, Paul. On Mon, Sep 3, 2018 at 4:06 PM <fo...@univ-mlv.fr> wrote: > > > ------------------------------ > > *De: *"paulk" <pa...@asert.com.au> > *À: *"Remi Forax" <fo...@univ-mlv.fr> > *Cc: *"dev" <dev@groovy.apache.org> > *Envoyé: *Lundi 3 Septembre 2018 00:26:27 > *Objet: *Re: Support of Java 11 using ASM 6.2.1 and ASM 7 > > Hi, > Groovy 3.x snapshots use ASM7_EXPERIMENTAL. For Groovy 2.5.x, the current > plan (subject to change) is that we were planning on waiting for the ASM 7 > release and doing a new release after that. > > > in that case, perhaps Groovy 2.5.x should reject classfiles V11 and V12 > with an error message instead of having ASM throwing an > UnsupportedOperationException later, at least until ASM 7 is integrated > > > But, I must admit that I haven't really had a chance to look at the full > implications of using ASM7_EXPERIMENTAL. If we added that already in 2.5.x > but don't use any of the new features, is there any backward compatibility > issues when we then switched to ASM7? > > > Currently no, just replacing ASM7_EXPERIMENTAL by ASM7 is enough and given > that Java 11 is already in release candidate phase, we do not expect any > compatibility issues. > Also, there is no bytecode changes planned for 12 so the latest release of > ASM 6 (6.2.1) also allows Java 12 bytecode. > > > Cheers, Paul. > > > regards, > Rémi > > > > On Mon, Sep 3, 2018 at 2:06 AM Remi Forax <fo...@univ-mlv.fr> wrote: > >> Forget to ling to the related bug: >> https://issues.apache.org/jira/browse/GROOVY-8727 >> >> cheers, >> Rémi >> >> ----- Mail original ----- >> > De: "Remi Forax" <fo...@univ-mlv.fr> >> > À: "dev" <dev@groovy.apache.org> >> > Envoyé: Dimanche 2 Septembre 2018 18:04:12 >> > Objet: Support of Java 11 using ASM 6.2.1 and ASM 7 >> >> > Hi all, >> > Java 11 introduces several new forward incompatible features* in the >> class file. >> > >> > Currently javac only uses one of them, nestmates [1], which allows to >> declare >> > that several classes are part of the same nest thus allow access to >> private >> > members in between them, obviously changing the semantics of the >> private access >> > for invokevirtual (see [2] if you want more info) if not forward >> compatible. >> > >> > Even if NestMates are declared as class attributes, we have decided in >> ASM to >> > not ignore them if there are present in the bytecode (a class compiled >> with >> > javac 11) because this change for the VM is far from innocuous, if you >> scrap >> > those attributes, you get IllegalAccessError laters, so if you use a >> > ClassVisitor configured with the ASM6 API and ASM see a nestmate related >> > attributes (NestHost or NestMembers) it will fail with an >> > UnsuportedOperationException. >> > >> > Moreover, even if the support of Java 11 will came with ASM7, we have >> decided to >> > introduce a new experimental API version (currently ASM7_EXPERIMENTAL) >> which >> > let you parse Java 11 using ASM6 (6.2+), so if you want your favorite >> language >> > to support Java 11, all visitors need to be upgraded to the api version >> > ASM7_EXPERIMENTAL. >> > You can also decide as before to wait until we release ASM7 (the first >> week end >> > after the release of Java as usual, so at the end of September) to use >> the ASM7 >> > api. >> > >> > regards, >> > Rémi >> > >> > * see also Constant Dynamic (http://openjdk.java.net/jeps/309) and >> Preview >> > Feature (http://openjdk.java.net/jeps/12) >> > >> > [1] http://openjdk.java.net/jeps/181 >> > [2] >> https://youtu.be/-k_IicifbxQ?list=PLX8CzqL3ArzVnxC6PYxMlngEMv3W1pIkn >> > >
