RFR (S): 7178667: ALT_EXPORT_PATH does not export server jvm on macosx

[David Holmes]

On 27/06/2012 6:13 PM, Staffan Larsen wrote:
> Can I have a Review for this change, please?

Ok. :)

David

> The very simple fix is here:
> http://cr.openjdk.java.net/~sla/7178667/webrev.02/
>
> Thanks,
> /Staffan
>
> On 25 jun 2012, at 10:36, Staffan Larsen wrote:
>
>>
>>>>> So, it sounds like when you rebuilt, everything was built into
>>>>> jre/lib/i386 and jre/lib/amd64, but never combined (or, in this
>>>>> case, just copied) into jre/lib, and therefore not found.
>>>>
>>>> Yes. Or rather, only the client jvm was combined, but the client jvm
>>>> isn't copied into the j2sdk-image on mac, so nothing was copied.
>>>
>>> Which begs the question: if we only build 64-bit on OSX then how/why
>>> is client being built in the first place?
>>
>> I should have said: "only the client jvm was _attempted_ to be
>> combined". In fact, the client does not exist, but the universalize
>> makefiles are written to handle client if it did exist.
>>
>> So what happened was:
>> - the product jvm was built
>> - it was copied to the import jdk (into jre/lib/amd64/server/) by the
>> generic_export target
>> - the universalize makefile tried to take the client jvm and
>> universalize it into jre/lib/client/ (notice that there is no amd64
>> directory level on mac)
>> - the universalize makefile removes all {amd64,i386} directories
>>
>> What should have happened:
>> - the product jvm was built
>> - it was copied to the import jdk (into jre/lib/amd64/server/) by the
>> generic_export target
>> - the universalize makefile makes a universal binary of any existing
>> jvms (client or server)
>> - the universalize makefile copies these jvms into jre/lib/{server,client}
>> - the universalize makefile removes all {amd64,i386} directories
>>
>> But because the targets weren't .PHONY, the third step above failed.
>>
>> I hope that explains the problem in more detail. Who wants to be put
>> down as reviewer?
>>
>> Thanks,
>> /Staffan
>>
>