Closed mpkorstanje closed 10 months ago
@xeger any clue?
@mpkorstanje, appears to be an open issue with tree-sitter-cli 0.20.8
(which we use), whereby parser generation varies across operating systems - tree-sitter/tree-sitter#2755.
This is potentially what's happening in our case due to files not being generated/found.
If that's the case, other than awaiting a fix, rolling back to 0.20.7
and preventing automerge should hopefully resolve the issue for the time being.
Edit: Will try to validate this assumption
Argh; this is the second time Tree sitter has had an obscure wasm bug. Suggest testing the rolled back build against the vscode plugin to ensure the electron wasm linking issue is not present in the older Tree sitter artifacts.
Raised a PR that resolves the issue. Caused by macos-latest
updating Python to 3.12 - with a breaking change in our pinned version of node-gyp
due to removal of distutils
from the standard library. The tree-sitter language bindings subsequently failed to build on installation and the root-cause of failure was suppressed due to the low verbosity of the logs and as they are specified as optional dependencies - which allows the run to continue to the compilation script even if failures are present.
Varying output was a result of our script asynchronously attempting compilation of each of the tree-sitter language bindings - and the failure observed being the first process that finished.
👓 What did you see?
Builds are failing intermittently when compiling some tree sitter modules, e.g:
https://github.com/cucumber/language-service/actions/runs/6752791276/job/18358568725
https://github.com/cucumber/language-service/actions/runs/7152098907/job/19477061875
📚 Any additional context?
I'm not familiar with this stack.