* Fix import mapping to work with built JS files
This allows us to ship a Docker image that runs using `node` directly,
rather than via `tsx`.
* Remove use of top-level `await` expressions
These were only used in tests and weren't critical. Switch away so we
have the optional to convert to CommonJS if we need.
* Switch all imports to use relative paths instead of ~ shortcut
The ~ shortcut doesn't work in some contexts, which makes our lives more
difficult when trying to output ESM.
Since we were already using relative paths in some part of the codebase
and not others, just switch to using relative paths everywhere for
simplicity.
* Switch imports to append .js extension
And update Jest configuration to work with this extension.
* Build single image instead of two separate ones
We originally created the other image so that we could build an image
using only packages published to NPM, but this proved problematic when
wanting to test those changes without publishing.
Since there are more situations where we'd want to ship an image using
unpublished packages (i.e. for testing) remove the "public" image and
update our "testing" image to also be used as the public image. However,
a key difference is that the testing image will now run compiled JS
instead of using a TypeScript interpreter.
* rolling up changes for links FIP
* typeToSetPostfix
* pr feedback
* consistency to avoid js quirks
* add versioning logic and update tests
* include version check in mergeMessages
* update protobuf comment to reflect nit
* added changeset
* code coverage
* chore: fix typo when deploying on mainnet
`FC_NETOWKR_ID` on the mainnet should be 1 rather than 2.
* chore: update peer id because of lose identity file
We had this implemented incorrectly. We need to error once it meets or
exceeds the limit, since you can only store (2^N - 1) for a binary value
with N bits.
* fix(eth): add stronger retry logic to eth events provider
* test has sporadic failure on CI, appears to be due to unresolved promise, perhaps this will fix
* alter retry provider to throw on retry limit
* more slight adjustments, makes concurrent tests stop freaking out
* add http status code to enforce retry
* slight changes to increase code coverage
* avoid clobbering globals
* await on emit
* finally hunted down RC, it's a timer, deep in subscriber
* ignoring mock utils
* exponential backoff + faster tests
We want to make it easier for developers to get started with hubs.
Add a separate optimized Docker image which we'll distribute for both
AMD and ARM architectures.
For now the publishing process is manual, but once hubs are fully open
(no peering allowlists) we'll be able to auto-publish with each hubble
release on NPM.
This dependency is only required for dev, not production. It's used by
Hubble when generating fake data, however, so move it to a production
dependency for Hubble.
With more Macs running ARM via Apple Silicon, the significantly slow
package installation was problematic. Fix by using our fork which
includes the prebuilt binary to reduce install time.
Fetch data in batches and execute jobs concurrently. This reduces the
time estimate from ~4 hours to under 1 hour on my machine.
We could probably make this even faster by implementing a
`getAllMessages` endpoint on the hubs.
Instead of requiring the user to run `yarn install` followed by `yarn
start`, we simply have them run `docker compose up` to start both the
app and Postgres together. This is much easier and reduces the need for
them to install anything else besides Docker. It also makes cleaning up
the example easier.
