RFR: 8234185: Cleanup usage of canonicalize function between libjava, hotspot and libinstrument

serguei.spitsyn at oracle.com serguei.spitsyn at oracle.com
Wed Nov 20 02:20:32 UTC 2019


Hi Christoph,

The fix looks good to me.
I'd recommend to run the jdk_instrument and vmTestbase_nsk_jvmti tests.
Also, it can be safe to run:
   open/test/hotspot/jtreg/serviceability/jvmti
   open/test/hotspot/jtreg/runtime/cds/appcds
   open/test/hotspot/jtreg/runtime/BootClassAppendProp

Thanks,
Serguei

On 11/14/19 07:37, Langer, Christoph wrote:
> Hi,
>
> please review this cleanup change regarding function "canonicalize" of libjava.
>
> Bug: https://bugs.openjdk.java.net/browse/JDK-8234185
> Webrev: http://cr.openjdk.java.net/~clanger/webrevs/8234185.0/
>
>
> The goal is to cleanup how this function is defined and used. One thing is, that there was an unnecessary wrapper function "Canonicalize" in jni_util.c. It wrapped the call to "canonicalize". We can get rid of this wrapper. Unfortunately, it is not possible to just export "canonicalize" since this will conflict with a method signature from the math library, at least on modern Linuxes. So I decided to call the method JDK_Canonicalize and will correctly define it in jdk_util.h which can be included everywhere.
>
>
>
> Hotspot's classloader.cpp will dynamically resolve this method, so I add a local declaration of the function pointer in there.
>
>
>
> This change shall be predecessor of JDK-8223261, where a review was already started here: https://mail.openjdk.java.net/pipermail/core-libs-dev/2019-November/063398.html
>
> Thanks
> Christoph
>



More information about the hotspot-runtime-dev mailing list