Use GetExtensionEvents to derive extension event numbers #545
Conversation
|
GetExtensionEvents JVMTI spec: https://download.java.net/java/early_access/jdk19/docs/specs/jvmti.html#GetExtensionEvents Ports: |
...otspot/jtreg/serviceability/jvmti/vthread/BreakpointInYieldTest/libBreakpointInYieldTest.cpp
Outdated
Show resolved
Hide resolved
|
Why is this necessary at all? We go to some effort in jvmti.h.m4 to (apparently) match the RI event numbers. Perhaps that's where the problem lies? |
@gacholio mentioned that the current behaviour is invalid and not supported by the JVMTI specification. More details can be found in the comments surrounding eclipse-openj9/openj9#16501 (comment). |
I'm not saying it isn't questionable practice to hard-code those constants, but I expect they will be constants, and others will do likewise, so we're only hurting ourselves by playing hardball here. Hopefully they'll be defined in |
|
Extension events are by definition not in the header file since they are not expected to be the same between VM implementations. There is a specified way to get the numbers from the data returned from the extension query. |
|
Ok got it. Thanks. |
|
Once we settle on a solution here, we should be proposing the fix upstream. |
babsingh
force-pushed
the
fix_exteventnumbers
branch
from
de6aa95 to
db6f8df
Compare
by upstream, do we mean https://github.com/openjdk/jdk? |
babsingh
force-pushed
the
fix_exteventnumbers
branch
from
db6f8df to
dda36cd
Compare
|
As this is throw-away code, the most likely way to get changes accepted upstream would be to make the smallest change possible, not the most elegant. |
Yes. |
|
I think this should wait for the resolution of #531 (comment). |
Currently, extension event numbers are hard-coded in the tests. This is invalid since the extension event numbers can vary between implementations. Instead, GetExtensionEvents should be used to derive the extension event numbers. Related: eclipse-openj9/openj9#16501 Related: eclipse-openj9/openj9#16167 Signed-off-by: Babneet Singh <[email protected]>
babsingh
force-pushed
the
fix_exteventnumbers
branch
from
dda36cd to
d3f3782
Compare
None of the files in this PR are updated in #531 (comment). There is no noticeable dependency. I don't see a reason to wait. |
The changes here appear to be the same as in ibmruntimes/openj9-openjdk-jdk20#13; once I'm satisfied with the latter, I think it should be safe to merge this. |
Currently, extension event numbers are hard-coded in the tests. This is
invalid since the extension event numbers can vary between
implementations. Instead, GetExtensionEvents should be used to derive
the extension event numbers.
Related: eclipse-openj9/openj9#16501
Related: eclipse-openj9/openj9#16167
Signed-off-by: Babneet Singh [email protected]