RFR: 8282608: RawNativeLibraryImpl can't be passed to NativeLibraries::findEntry0

Mandy Chung mchung at openjdk.java.net
Thu Mar 3 22:39:59 UTC 2022


On Thu, 3 Mar 2022 22:14:15 GMT, Maurizio Cimadamore <mcimadamore at openjdk.org> wrote:

>> This fixes a bug in JDK-8282515 that `NativeLibraries::findEntry0` reads the handle from the given native library instance with the field ID for `NativeLibraryImpl::handle` which is wrong for `RawNativeLibraryImpl`.  The `NativeLibraries::findEntry0` now takes the handle of the library instead and that can be used to look up symbols in the native library loaded by System::loadLibrary and also RawNativeLibraries.
>> 
>> Tier 1-4 testing is in progress.
>
> src/java.base/share/classes/jdk/internal/loader/RawNativeLibraries.java line 38:
> 
>> 36: import java.util.concurrent.ConcurrentHashMap;
>> 37: 
>> 38: import static jdk.internal.loader.NativeLibraries.findEntry0;
> 
> We have `NativeLibraries` and `RawNativeLibraries` which are unrelated classes. Then we have `NativeLibraryImpl` and `RawNativeLibraryImpl` which shared a common super type (`NativeLibrary`). I find it mildly asymmetric to put the `find0` native method in `NativeLibraries` since that's not the supertype of `RawNativeLibraries`. Is there common code between the two XYZLibraries classes that can be shared? If so, maybe `find0` can go in that common supertype - if not, maybe it's better to put `find0` in `NativeLibrary` itself?

I agree it's asymmetric to put the native method in `NativeLibraries`.   `NativeLibrary` would be a better place for `findEntry0`.  I will also turn `NativeLibrary` to an abstract class to make `findEntry0` is package-private static method.

-------------

PR: https://git.openjdk.java.net/jdk/pull/7684


More information about the core-libs-dev mailing list