RFR: 8245717: VM option "-XX:EnableJVMCIProduct" could not be repetitively enabled

[David Holmes]

Hi,

On 26/05/2020 4:57 pm, Xiaohong Gong wrote:
> Hi,
> 
> Could you please help to review this simple patch? It fixes the issue that JVM crashes
> in debug mode when the vm option "-XX:EnableJVMCIProduct" is enabled repetitively.
> 
> JBS: https://bugs.openjdk.java.net/browse/JDK-8245717
> Webrev: http://cr.openjdk.java.net/~xgong/rfr/8245717/webrev.00/
> 
> Repetitively enabling the vm option "-XX:+EnableJVMCIProduct" in the command
> line makes the assertion fail in debug mode: "assert(is_experimental(), sanity)".
> 
> It happens when the VM iteratively parses the options from command line. When the
> matched option is "-XX:+EnableJVMCIProduct", the original experimental JVMCI flags
> will be converted to product mode, with the above assertion before it. So if all
> the JVMCI flags have been converted to the product mode at the first time parsing
> "-XX:+EnableJVMCIProduct", the assertion will fail at the second time it is parsed.
> 
> A simple fix is to just ignoring the conversion if this option has been parsed.

Seems a reasonable approach given the already complex handling of this flag.

> Testing:
> Tested jtreg hotspot::hotspot_all_no_apps,jdk::jdk_core,langtools::tier1
> and jcstress:tests-custom, and all tests pass without new failure.

I think adding a regression test in 
./compiler/jvmci/TestEnableJVMCIProduct.java would be appropriate.

Thanks,
David

> Thanks,
> Xiaohong Gong
>