Native Node Modules
For some features, the desktop version of Element can make use of native Node modules. These allow Element to integrate with the desktop in ways that a browser cannot.
While native modules enable powerful new features, they must be complied for each operating system. For official Element releases, we will always build these modules from source to ensure we can trust the compiled output. In the future, we may offer a pre-compiled path for those who want to use these features in a custom build of Element without installing the various build tools required.
The process is automated by vector-im/element-builder when releasing.
Building
Install the pre-requisites for your system:
- Windows pre-requisites
- Linux: TODO
- OS X: TODO
Then optionally, add seshat and dependencies to support search in E2E rooms.
Then, to build for an architecture selected automatically based on your system (recommended), run:
yarn run build:native
If you need to build for a specific architecture, see here.
Adding Seshat for search in E2E encrypted rooms
Seshat is a native Node module that adds support for local event indexing and full text search in E2E encrypted rooms.
Since Seshat is written in Rust, the Rust compiler and related tools need to be installed before installing Seshat itself. To install Rust please consult the official Rust documentation.
Seshat also depends on the SQLCipher library to store its data in encrypted form on disk. You'll need to install it via your OS package manager.
After installing the Rust compiler and SQLCipher, Seshat support can be added using yarn at the root of this project:
yarn add matrix-seshat
You will have to rebuild the native libraries against electron's version
of node rather than your system node, using the electron-build-env
tool.
This is also needed to when pulling in changes to Seshat using yarn link
.
yarn add electron-build-env
Recompiling Seshat itself can be done like so:
yarn run electron-build-env -- --electron 6.1.1 -- neon build matrix-seshat --release
Please make sure to include all the --
as well as the --release
command line
switch at the end. Modify your electron version accordingly depending on the
version that is installed on your system.
After this is done the Electron version of Element can be run from the main folder as usual using:
yarn start
Statically linking libsqlcipher
On Windows & macOS we always statically link libsqlcipher for it is not generally available.
On Linux by default we will use a system package, on debian & ubuntu this is libsqlcipher0
,
but this is problematic for some other packages, and we found that it may crashes for unknown reasons.
By including SQLCIPHER_BUNDLED=1
in the build environment, the build scripts will fully statically
link sqlcipher, including a static build of OpenSSL.
More info can be found at https://github.com/matrix-org/seshat/issues/102 and https://github.com/vector-im/element-web/issues/20926.
Compiling for specific architectures
macOS
On macOS, you can build universal native modules too:
yarn run build:native:universal
...or you can build for a specific architecture:
yarn run build:native --target x86_64-apple-darwin
or
yarn run build:native --target aarch64-apple-darwin
You'll then need to create a built bundle with the same architecture. To bundle a universal build for macOS, run:
yarn run build:universal
Windows
If you're on Windows, you can choose to build specifically for 32 or 64 bit:
yarn run build:32
or
yarn run build:64
Cross compiling
Compiling a module for a particular operating system (Linux/macOS/Windows) needs to be done on that operating system. Cross-compiling from a host OS for a different target OS may be possible, but we don't support this flow with Element dependencies at this time.
Switching between architectures
The native module build system keeps the different architectures
separate, so you can keep native modules for several architectures at the same
time and switch which are active using a yarn run hak copy
command, passing
the appropriate architectures. This will error if you haven't yet built those
architectures. eg:
yarn run build:native --target x86_64-apple-darwin
# We've now built & linked into place native modules for Intel
yarn run build:native --target aarch64-apple-darwin
# We've now built Apple Silicon modules too, and linked them into place as the active ones
yarn run hak copy --target x86_64-apple-darwin
# We've now switched back to our Intel modules
yarn run hak copy --target x86_64-apple-darwin --target aarch64-apple-darwin
# Now our native modules are universal x86_64+aarch64 binaries
The current set of native modules are stored in .hak/hakModules
,
so you can use this to check what architecture is currently in place, eg:
$ lipo -info .hak/hakModules/keytar/build/Release/keytar.node
Architectures in the fat file: .hak/hakModules/keytar/build/Release/keytar.node are: x86_64 arm64