RFR: 8273801: Handle VMTYPE for "core" VM variant

[David Holmes]

On 20/09/2021 9:43 pm, Magnus Ihse Bursie wrote:
> On Thu, 16 Sep 2021 12:02:38 GMT, David Holmes <dholmes at openjdk.org> wrote:
> 
>>> Building with --with-jvm-variants=core currently produces a binary that replies an odd version:
>>>
>>>
>>> $ build/linux-x86_64-core-fastdebug/images/jdk/bin/java -version
>>> openjdk version "18-internal" 2022-03-15
>>> OpenJDK Runtime Environment (fastdebug build 18-internal+0-adhoc.shade.jdk)
>>> OpenJDK 64-Bit  VM (fastdebug build 18-internal+0-adhoc.shade.jdk, interpreted mode)
>>>
>>>
>>> It should actually say "OpenJDK 64-Bit Core VM", I think. Build just misses a simple definition of `VMTYPE` for core variant.
>>>
>>> After the patch:
>>>
>>>
>>> $ build/linux-x86_64-core-fastdebug/images/jdk/bin/java -client -version
>>> openjdk version "18-internal" 2022-03-15
>>> OpenJDK Runtime Environment (fastdebug build 18-internal+0-adhoc.shade.jdk)
>>> OpenJDK 64-Bit Core VM (fastdebug build 18-internal+0-adhoc.shade.jdk, interpreted mode)
>>
>> Begs the question ... if there is no C1 and no C2 then what is the name of the directory where libjvm.so lives?
> 
> @dholmes-ora If you are building `server`, then it is called `server` regardless of what JVM features you enable or disable.

Doh! Of course. The variant determines the name.

Thanks,
David

> The way the build system works is that a "JVM variant" is a set of enabled JVM features with a given name. You can roll your own JVM variant by starting with `custom`, which has an empty initial set of features enabled. (But currently, you cannot change the name from the configure command line.)
> 
> If core is built only when porting to new platforms, we could emulate that behavior by allowing the user to change the JVM variant name as well, like `configure --with-jvm-variant=server --with-jvm-features=-compiler1,-compiler2 --with-jvm-name=core` or something like that. (And just generate VMTYPE dynamically using Sentence Case). I think this can be seriously considered. But I suspect I should try to dug out from my mailbox archives whom it was last told me they needed/wanted core.
> 
> -------------
> 
> PR: https://git.openjdk.java.net/jdk/pull/5525
>