RFR(M): 8223660: jtreg: Decouple Unsafe from RTM tests

Doerr, Martin martin.doerr at sap.com
Fri May 24 09:33:39 UTC 2019


Hi Gustavo,

looks good to me. Tested on AIX. The previously failing tests have passed.

Best regards,
Martin


> -----Original Message-----
> From: Gustavo Romero <gromero at linux.vnet.ibm.com>
> Sent: Mittwoch, 22. Mai 2019 21:26
> To: hotspot-compiler-dev at openjdk.java.net
> Cc: Doerr, Martin <martin.doerr at sap.com>; Lindenmaier, Goetz
> <goetz.lindenmaier at sap.com>; vladimir.kozlov at oracle.com
> Subject: RFR(M): 8223660: jtreg: Decouple Unsafe from RTM tests
> 
> Hi,
> 
> Could I get reviews for the following change please?
> 
> Bug   : https://bugs.openjdk.java.net/browse/JDK-8223660
> Webrev: http://cr.openjdk.java.net/~gromero/8223660/v1/
> 
> It removes from the RTM jtreg tests the use of Unsafe native methods as
> abort provokers in a transaction.
> 
> Relying on Unsafe native methods to abort a transaction makes the RTM test
> brittle because an Unsafe native method can be converted to non-native at
> any time, breaking the RTM tests. This is the second time it happens.
> 
> This change removes the use of Unsafe native methods (currently, pageSize()
> is used, but is not native anymore) and adds an isolated native method in
> library libXAbortProvoker.so for the XAbortProvoker class that will be used
> to abort transactions in the RTM tests, turning the RTM jtreg tests more
> self-contained and so less brittle.
> 
> I tested the change on x86_64 and PPC64, w/ RTM/HTM CPU feature
> available.
> 
> Thank you!
> 
> Best regards,
> Gustavo


More information about the hotspot-compiler-dev mailing list