How to dispatch the same method call on different platforms?

[tison]

It even helps distributions ... I found that C#/NuGet may build a
conventional way to distribute multi-platform dylibs [1].

But in the Java world, artifacts are released to Maven repositories. It
uses another incomplete slang to determine platform [2]. And user-specific
classifier for distinguishing dylibs (many many slangs).

System.loadLibrary doesn't load some conventional paths (or I don't know),
make Java libs based on dylibs have to write their own native loader [3]:
another place where many tribal rules various among projects.

Best,
tison.

[1] https://github.com/confluentinc/librdkafka/wiki/librdkafka.redist
[2] https://maven.apache.org/enforcer/enforcer-rules/requireOS.html
[3]
https://github.com/apache/opendal/blob/main/bindings/java/src/main/java/org/apache/opendal/NativeLibrary.java


Christoph Läubrich <laeubi at laeubi-soft.de> 于2024年2月17日周六 15:15写道：

> It would really great if the JVM (maybe as part of FFM) can offer a way
> to get a more semantic wan to query for os/architecture than system
> properties with varying names.
>
> Especially as the JVM is a native binary it should know all the details ;-)
>
> In OSGi there is a namespace one can use to filter on [1].
>
> It even defines an Alias list[2], just to give an impression about the
> mess one need to handle today with all the different names (what is
> maybe even incomplete as of today).
>
> SWT (who has different platforms/arch implementations) has a platform
> specific java class that defines a constant with a well known value that
> can be used at runtime to find that out [3]
>
>
> If then jextract would be able to get a list of alternative headers for
> different platforms... just dreaming ;-)
>
> best Christoph
>
> [1]
>
> https://docs.osgi.org/specification/osgi.core/8.0.0/framework.namespaces.html#framework.namespaces.osgi.native
> [2] https://docs.osgi.org/reference/osnames.html
> [3]
>
> https://github.com/eclipse-platform/eclipse.platform.swt/blob/master/bundles/org.eclipse.swt/Eclipse%20SWT%20PI/gtk/org/eclipse/swt/internal/Platform.java
> [4]
>
> https://github.com/eclipse-platform/eclipse.platform.swt/blob/master/bundles/org.eclipse.swt/Eclipse%20SWT%20PI/win32/org/eclipse/swt/internal/Platform.java
>
>
> Am 17.02.24 um 07:53 schrieb tison:
> > For example, getpwnam returns passwd struct, but the struct layout is
> > various on different platform, making the code snippet below unportable:
> >
> >          final Linker linker = Linker.nativeLinker();
> >          final SymbolLookup lookup = linker.defaultLookup();
> >          final MemorySegment getpwnam =
> > lookup.find("getpwnam").orElseThrow();
> >          final MethodHandle getpwnamFn = linker.downcallHandle(getpwnam,
> > FunctionDescriptor.of(ValueLayout.ADDRESS, ValueLayout.ADDRESS));
> >          final StructLayout passwdLayout = MemoryLayout.structLayout(
> >                  ValueLayout.ADDRESS.withName("pw_name"),
> >                  ValueLayout.ADDRESS.withName("pw_passwd"),
> >                  ValueLayout.JAVA_INT.withName("pw_uid"),
> >                  ValueLayout.JAVA_INT.withName("pw_gid"),
> >                  ValueLayout.JAVA_LONG.withName("pw_change"),
> >                  ValueLayout.ADDRESS.withName("pw_class"),
> >                  ValueLayout.ADDRESS.withName("pw_gecos"),
> >                  ValueLayout.ADDRESS.withName("pw_dir"),
> >                  ValueLayout.ADDRESS.withName("pw_shell"),
> >                  ValueLayout.JAVA_LONG.withName("pw_expire"),
> >                  ValueLayout.JAVA_INT.withName("pw_fields"));
> >          try (final Arena arena = Arena.ofConfined()) {
> >              final MemorySegment username =
> arena.allocateUtf8String(user);
> >              final MemorySegment passwd = ((MemorySegment)
> > getpwnamFn.invoke(username)).reinterpret(Long.MAX_VALUE);
> >              final MemorySegment dir = passwd.get(
> >                      ValueLayout.ADDRESS,
> >
> >
> passwdLayout.byteOffset(MemoryLayout.PathElement.groupElement("pw_dir")));
> >
> > System.out.println(dir.reinterpret(Long.MAX_VALUE).getUtf8String(0));
> >          }
> >
> > This code snippet only works on macOS because the layout differs on
> > other platforms.
> >
> > In Rust, we can use #[target(os = ..)] to switch the manner, and in
> > Java, perhaps we can use inheritance or interfaces, but it still lacks:
> >
> > * A unified way to determine current os, arch, toolchain, etc. I made
> > [1] that can help but it's no more than another incomplete slang.
> > * A compile-time dispatch decision. Maybe with static initialization
> > block it can helps by generating 'static final'
> > MethodHandle/VarHandle. I don't know.
> >
> > Best,
> > tison.
> >
> > [1] github.com/tisonspieces/os-detector
> > <http://github.com/tisonspieces/os-detector>
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/panama-dev/attachments/20240217/717bc51e/attachment-0001.htm>