Share package.json scripts across packages #239
Conversation
… wasm from bb and circuits.js
|
|
The script then would be able to copy scripts from a json to a set of packages given at command line. I guess it'd be nice if we had real inheritance, but this feels like a little too scrappy |
What I'm worried about that solution is that it's easy for packages to get out of sync, for a dev to foget to include a set of packages, etc. After all, the issue with the missing prepare step was only uncovered by pushing the same config all over the place. I agree that this feels scrappy, but manipulating the file directly is the only option given the lack of extensibility at the package.json level. Maybe adding a mechanism for individual packages to override a specific script would help? Or would it add to the scrappiness? |
|
I'm having trouble thinking of a better solution besides having some 'yarn common build' instead of yarn build, and routing everything through a common command. OK with this though for lack of a much better idea |
* fix bug in native kernel - contract logic was being skipped * zero-initialize all abi and kernel structs. helper function to zero-init arrays of frs * better struct initialization in app circuits files * dont need to do as much in base utils since structs are sanely zero-initialized * use fill to initialize array of fields to zero * fix native private kernel, refactor private kernel tests, add real logic in test setup for computing function/contract tree roots/siblings * new hash functions
* fix bug in native kernel - contract logic was being skipped * zero-initialize all abi and kernel structs. helper function to zero-init arrays of frs * better struct initialization in app circuits files * dont need to do as much in base utils since structs are sanely zero-initialized * use fill to initialize array of fields to zero * fix native private kernel, refactor private kernel tests, add real logic in test setup for computing function/contract tree roots/siblings * new hash functions
* fix bug in native kernel - contract logic was being skipped * zero-initialize all abi and kernel structs. helper function to zero-init arrays of frs * better struct initialization in app circuits files * dont need to do as much in base utils since structs are sanely zero-initialized * use fill to initialize array of fields to zero * fix native private kernel, refactor private kernel tests, add real logic in test setup for computing function/contract tree roots/siblings * new hash functions
Sets up a shared
package.scripts.jsonat the project root, which gets injected into each individual package.json that extends from it (via aninheritsdirective) wheneveryarn prepareis run. Ended up using a custom script over https://github.com/microsoft/package-inherit since it didn't work with our workspace setup, seems unmaintained, and it was easier to write manually.Something funny that happened after setting this up: forcing all packages to have the same set of scripts revealed that circuits and bb.js didn't have a prepare step, so their entry in the build_manifest wasn't getting updated. However, their entries depend on non-packages (ie wasm build), which cannot be expressed from the package json. So I added a near-empty
yarn-project-base/package.json, which does depend on the wasm build in the build manifest, and added devDeps to it from circuits and bb js. This ensures we have a transitive dependency to wasm builds from those packages.Fixes #202