RFR: 8143155: Remove TraceRuntimeCalls, TraceJNICalls, and TraceJVMCalls rather than convert to UL (original) (raw)

Coleen Phillimore coleen.phillimore at oracle.com
Fri Nov 20 21:50:45 UTC 2015

This is great that there was extra cleanup. Thanks to David for noticing it.

On 11/20/15 3:03 PM, Rachel Protacio wrote:

Thank you, Dmitry and David!

Updated webrev: http://cr.openjdk.java.net/~rprotacio/8143155.01/ I reinstated the VMOptionWarning.java test cases, and in jvm.cpp deleted the unused wrappers. David, thanks for pointing that out. I took the liberty of deleting the %s/%d/etc. parts of the messages since they previously were printing those out literally (and not substituting the variables), which I think was definitely not intended. In the unlikely case that you think anyone was relying on those exact strings, let me know and I'll change them back?

The %s/%d were showing up literally in CountJVMCalls, not TraceJVMCalls. Either way, your change is good to remove it.

The VMOptionWarning.java test should be changed to use CountJVMCalls and CountJNICalls, or some other develop only option.

Thanks, Coleen

Thank you, Rachel On 11/19/2015 5:23 PM, David Holmes wrote: Hi Rachel,

Overall seems okay. One minor issue below. Thanks, David On 20/11/2015 5:39 AM, Rachel Protacio wrote: Hi,

Please review this change, which removes TraceRuntimeCalls, TraceJNICalls, and TraceJVMCalls options. The output from the options is excessive and useless, while there are entirely useful options available, i.e. CountRuntimeCalls, CountJNICalls, and CountJVMCalls. Open webrev: http://cr.openjdk.java.net/~rprotacio/8143155/ Bug: https://bugs.openjdk.java.net/browse/JDK-8143155 src/share/vm/prims/jvm.cpp These: #define JVMWrapper(arg1) JVMCountWrapper(arg1); ! #define JVMWrapper2(arg1, arg2) JVMCountWrapper(arg1); ! #define JVMWrapper3(arg1, arg2, arg3) JVMCountWrapper(arg1); ! #define JVMWrapper4(arg1, arg2, arg3, arg4) JVMCountWrapper(arg1); should reduce to just the single form now that only 1 arg is needed. And the 3/4 variants are unused anyway. That said something seems a bit broken here as we have things like: JVMWrapper2("JVMNativePath (%s)", path); which will expand to JVMCountWrapper("JVMNativePath (%s)") which doesn't really make sense to me. I guess it is harmless, but certainly looks odd. There are only 10 uses to fix up though :)

Thank you! Rachel

More information about the hotspot-runtime-dev mailing list