Since our last release in February, we’ve had ~150 commits to Lemmy. The biggest changes, as we’ll outline below, are a split of Lemmy’s user tables into federated and local tables, necessitating a
v3 of Lemmy’s API, federated moderation, i18n support in join.lemmy.ml, and lots of back-end cleanup.
Important: there are multiple breaking changes:
blocked_instanceshas changed, and you need to adjust your config file manually:
allowed_instances: ["ds9.lemmy.ml", "enterprise.lemmy.ml"], and only one of the
blocked_instancesblocks can be set.
v3, so all clients need to be updated: lemmy-js-client: 0.9.9 -> 0.10.0 .
If you’d like to make a DB backup before upgrading, follow this guide.
To upgrade your instance to
v0.10.0, simply follow the instructions in the documentation:
|v0.9.0 (Rust 1.47)||v0.10.0 (Rust 1.47)||v0.10.0 (Rust 1.51)|
Despite ongoing efforts to speed up compilation, it has actually gotten slower when comparing with the same Rust version. Only thanks to improvements in newer Rust versions has our build process gotten faster. This could be simply because we added more code, while Lemmy v0.9.0 had 22.4k lines of Rust, v0.10.0 has 23.8k (an increase of 6%).
v0.9.0 build graph:
v0.10.0 build graph:
We extracted the crates
lemmy_apub, respectively, and renamed
lemmy_api_common. In the second graph you can see how parts of the api and apub crates are now built nicely in parallel, speeding up builds on multi-core systems.
On the other hand, some crates have gotten much slower to compile, in particular
lemmy_db_queries (6.5s slower),
lemmy_apub (6.5s slower if we include
lemmy_db_views is quite slow, just as before.