Build error on Linux PPC64 platform

[Volker Simonis]

Hi Neil,

thanks again for the detailed explanation. But it's still strange, why
the same 'rmic' invocation succeeds on my machine but not on
Jonathan's.

As posted I use the following IBM JDK:

java version "1.7.0"
Java(TM) SE Runtime Environment (build pxp6470-20110827_01)
IBM J9 VM (build 2.6, JRE 1.7.0 Linux ppc64-64 20110810_88604 (JIT
enabled, AOT enabled)
J9VM - R26_Java726_GA_20110810_1208_B88592
JIT  - r11_20110810_20466
GC   - R26_Java726_GA_20110810_1208_B88592
J9CL - 20110810_88604)
JCL - 20110809_01 based on Oracle 7b147

I checked "rmicext.properties" from ibmorbtools.jar. It has the
followng entries:

generator.args=v1.1,vcompat,v1.2,iiop,idl,xprint,mnt

generator.class.default=sun.rmi.rmic.RMIGenerator

generator.class.v1.1=sun.rmi.rmic.RMIGenerator
generator.class.vcompat=sun.rmi.rmic.RMIGenerator
generator.class.v1.2=sun.rmi.rmic.RMIGenerator
generator.class.iiop=com.ibm.tools.rmic.iiop.StubGenerator
generator.class.idl=com.ibm.tools.rmic.iiop.IDLGenerator
generator.class.xprint=com.ibm.tools.rmic.iiop.PrintGenerator
generator.class.mnt=com.ibm.tools.rmic.iiop.MangledNameTableGenerator

So according to your explanation, Jonathan should get the same results
for running "<ibm_jdk>/bin/java -cp <ibm_jdk>/lib/tools.jar
sun.rmi.rmic.Main -d /tmp/rmic_output/ -iiop -v1.2 -standardPackage
javax.management.remote.rmi.RMIConnectionImpl" as I did. Or is there
something else which can influence the choice of the generator class.

As far as I can see, the only place in Oracle's JDK which parses the
"-standardPackage" option is in  sun.rmi.rmic.iiop.StubGenerator
(http://hg.openjdk.java.net/ppc-aix-port/jdk7u/corba/file/tip/src/share/classes/sun/rmi/rmic/iiop/StubGenerator.java)
but according to your explanation, my IBM JDK should choose
"generator.class.iiop=com.ibm.tools.rmic.iiop.StubGenerator" in that
case.  However if I do:

<ibm_jdk>/bin/java -showversion -verbose:class -cp
<ibm_jdk>/lib/tools.jar sun.rmi.rmic.Main -d /tmp/rmic_output/ -iiop
-v1.2 -standardPackage javax.management.remote.rmi.RMIConnectionImpl
2>&1 | grep iiop.StubGenerator
class load: sun.rmi.rmic.iiop.StubGenerator from:
file:/net/sapmnt.depot/tools/gen/linuxppc64/licenseware/jse/1.7.0/lib/tools.jar

I can see that it uses sun.rmi.rmic.iiop.StubGenerator instead. So
there must some other configuration magic (at least on my machine).

@Jonathan: could you please try the above command to check which
version of StubGenerator you get.

Regards,
Volker


On Thu, Jul 5, 2012 at 2:43 PM, Neil Richards <neil.richards at ngmr.net> wrote:
> On Thu, 2012-07-05 at 11:59 +0200, Volker Simonis wrote:
>> Hi Jonathan,
>>
>> that's strange. We speciall set the 'rmic' compiler to
>>
>> RMIC='$(ALT_BOOTDIR)/bin/java -cp $(ALT_BOOTDIR)/lib/tools.jar
>> sun.rmi.rmic.Main'
>>
>> in order to use the original, Oracle version of the compiler which
>> should understand this option.
>
> I fear that this is not altogether sufficient to avoid using IBM's
> version of the rmic compiler for IIOP-related artifact generation.
>
> rmic is the command for producing both Java Remote Method Protocol
> (JRMP) and Internet Inter-ORB Protocol (IIOP) artifacts.
> (Associated with IIOP generation, it also can produce IDL code and a
> "Mangled Name Table").
>
> Each of these types of generation is done by a separate "back-end" to
> rmic.
>
> The back-end implementation for each generator can be specified (usually
> by an SDK provider) using the properties file
> 'sun/rmi/rmic/resources/rmicext.properties' (and its associated
> locale-specific variants).
>
> (These properties settings are picked up by the common rmic front-end,
> using a call to ResourceBundle.getBundle().)
>
> These properties files (if they exist) tend to be held within a jar
> file, alongside the code for the backend implementations.
>
> In an IBM SDK, that code, as so that properties file, is held in
> lib/ibmorbtools.jar.
> (The IBM SDK's lib/tools.jar has a 'Class-Path:' entry in it, which
> chains onto ibmorbtools.jar).
>
> So, in all cases, the rmic command invocation calls sun.rmi.rmic.Main,
> but the settings in the rmicext.properties file controls what backend
> code gets used.
>
> (Note: This is different to the way the idlj command works, for example,
> where there is no common front-end).
>
> Hope this helps to explain why the current approach is not having the
> desired effect.
>
> Regards,
> Neil
>
> --
> Unless stated above:
> IBM email: neil_richards at uk.ibm.com
> IBM United Kingdom Limited - Registered in England and Wales with number 741598.
> Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU
>