This repo contains an autogeneration system to generate LDK bindings for garbage-collected languages, currently including Java and TypeScript. See below for the current status of the bindings.
-The auto-generated code does not yet contain documentation, however the API mirrors rust-lightning exactly, so all documentation at [docs.rs/lightning](https://docs.rs/lightning) is applicable. High-level documentation of the API can be found at [lightningdevkit.org](https://lightningdevkit.org).
+The auto-generated code contains copies of the Rust documentation, which can also be viewed at
+[docs.rs/lightning](https://docs.rs/lightning). High-level documentation of the API can be found at
+[lightningdevkit.org](https://lightningdevkit.org).
Building
========
A release build of the Java bindings library for Linux is available in git. Thus, the bindings should work as long as the `LD_LIBRARY_PATH` includes the top-level directory of this repository.
-To build the bindings locally, the bindings require some additional work which is still making its way upstream, for now it should be built against the [rust-lightning 2021-03-java-bindings-base branch on git.bitcoin.ninja](https://git.bitcoin.ninja/?p=rust-lightning;a=shortlog;h=refs/heads/2021-03-java-bindings-base). Check that branch out locally and run the `genbindings.sh` script in it to build the required binaries. Thereafter, in this repo, run the `genbindings.sh` script with the first argument pointing to the rust-lightning directory, the second the relevant JNI CFLAGS, the third argument set to `true` or `false` to indicate whether building in debug mode, and the third set to true or false to indicate if the bindings should be built with workarounds required for Android. JNI CFLAGS on debian are likely "-I/usr/lib/jvm/java-11-openjdk-amd64/include/ -I/usr/lib/jvm/java-11-openjdk-amd64/include/linux/". When running a program linking against the library in debug mode, LD_PRELOAD should likely include the relevant `libclang_rt.asan-platform.so` path.
+To build the bindings locally, the bindings require some additional work which is still making its
+way upstream, for now it should be built against the
+[rust-lightning 2021-03-java-bindings-base branch on git.bitcoin.ninja](https://git.bitcoin.ninja/?p=rust-lightning;a=shortlog;h=refs/heads/2021-03-java-bindings-base).
+Check that branch out locally as well as [ldk-c-bindings](https://github.com/lightningdevkit/ldk-c-bindings)
+and run the `genbindings.sh` script in ldk-c-bindings to build the required binaries. Thereafter,
+in this repo, run the `genbindings.sh` script with the first argument pointing to the ldk-c-bindings
+directory, the second the relevant JNI CFLAGS, the third argument set to `true` or `false` to
+indicate whether building in debug mode, and the third set to true or false to indicate if the
+bindings should be built with workarounds required for Android. JNI CFLAGS on debian are likely
+"-I/usr/lib/jvm/java-11-openjdk-amd64/include/ -I/usr/lib/jvm/java-11-openjdk-amd64/include/linux/".
+When running a program linking against the library in debug mode, LD_PRELOAD should likely include
+the relevant `libclang_rt.asan-platform.so` path.
Status
======
-The TypeScript Bindings are still in early development any generated code contains syntax errors.
-
-While the underlying library and C bindings are relatively mature, the Java bindings should be considered alpha quality. Around 98% of the Rust API surface is exposed (with one or two functions around peer feature sets still to be exposed), but some memory management issues may still appear. Specifically, because the Java bindings map between two very different memory models - Rust's strict ownership model and Java's reference cloning and garbage collection - a lot of work occurs in the background to keep the Java GC informed of Rust ownership semantics.
-
-There are some known memory leaks, which are relatively modest in the existing test suite, but which may be less moderate in certain usages. The debug-mode build includes memory leak tracking and will print all loose objects when the program exists, though without calls to `System.gc(); System.runFinalization();` immediately before exit there will likely be many false positives. While it will require some complicated usage, there are likely some use-after-free or unkonwn-free bugs remaining. The debug-mode build links LLVM address sanitizer and will print diagnostic information in case of such issues.
-
-The only known issue resulting in a use-after-free bug requires custom a custom ChannelKeys instance created as a part of a new channel. After the channel is created, the ChannelKeys object will not be freed while the parent ChannelManager exists, however if the ChannelManager is garbage collected while a ChannelMonitor object which is associated with the same channel exists, a use-after-free bug may occur. This issue should be relatively rare as uses where a ChannelManager is removed while associated ChannelMonitors exist is not anticipated.
+The TypeScript Bindings are still in early development and generated code contains syntax errors.
+
+While the underlying library and C bindings are relatively mature, the Java bindings should be
+considered alpha quality and some memory management issues may still appear. Specifically, because
+the Java bindings map between two very different memory models - Rust's strict ownership model and
+Java's reference cloning and garbage collection - a lot of work occurs in the background to keep
+the Java GC informed of Rust ownership semantics.
+
+There are some known memory leaks, which are relatively modest in the existing test suite (around
+1MB for 128 node constructions and 64 full channel constructions and payments sent), but which may
+be less moderate in certain usages. The debug-mode build includes memory leak tracking and will
+print all loose objects when the program exists, though without calls to
+`System.gc(); System.runFinalization();` immediately before exit there will likely be many false
+positives. While it will require some complicated usage, there are likely some use-after-free or
+unkonwn-free bugs remaining. The debug-mode build links LLVM address sanitizer and will print
+diagnostic information in case of such issues.
+
+The only known issue resulting in a use-after-free bug requires custom a custom ChannelKeys instance
+created as a part of a new channel. After the channel is created, the ChannelKeys object will not
+be freed while the parent ChannelManager exists, however if the ChannelManager is garbage collected
+while a ChannelMonitor object which is associated with the same channel exists, a use-after-free bug
+may occur. This issue should be relatively rare as uses where a ChannelManager is removed while
+associated ChannelMonitors exist is not anticipated.
projects / ldk-java / commitdiff