RFR(S) : 8206117 : failed to get JDK properties for JVM w/o JVMCI
igor.ignatyev at oracle.com
Fri Jun 29 19:51:09 UTC 2018
I ain't sure setting 'false' to nonexistent flags would the right choice in all cases, and I don't want real 'false' and 'false' from nonexistent flags to be mixed up.
> On Jun 29, 2018, at 12:34 PM, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:
> Why not to check for 'null' in VMProps.java and assign 'false' value in such case?
> On 6/29/18 12:25 PM, Igor Ignatyev wrote:
>>> 39 lines changed: 0 ins; 0 del; 39 mod;
>> Hi all,
>> could you please review this small fix? if JVM has been build w/o JVMCI, it won't have EnableJVMCI flag, as a result jtreg-ext/requires/VMProps will throw NPE trying to set 'vm.opt.final.EnableJVMC' property. w/ this fix, we will get 'null' value for nonexistent flags. compiler/graalunit/graalunit/generateTests.sh has been updated to generate tests which can handle such values, compiler/graalunit/ tests have been regenerated.
>> webrev: http://cr.openjdk.java.net/~iignatyev//8204517/webrev.00/index.html
>> JBS: https://bugs.openjdk.java.net/browse/JDK-8206117
>> testing: compiler/graalunit tests on builds w/ and w/o JVMCI
>> -- Igor
More information about the hotspot-dev