feat(cli): concurrent builds of the same grammar on different locations on disk#5174
Conversation
WillLillis
left a comment
There was a problem hiding this comment.
Thanks for digging into this! I have a few suggestions/questions.
a656792 to
a3f0f6b
Compare
|
But on a more general note, I am not quite sure why the locking is necessary: to protect output path or the parser path ? If the former, then this PR addresses it, but if it's the latter, shouldn't we hash |
My understanding is that the locking is protecting the output path. This PR addresses that issue and (presumably) fixes the problem you were running into with tsdl, so I think it's good as is :) |
Instead of panicking somehere else. This happens on concurrent builds of the the same grammar.
a3f0f6b to
310d0fb
Compare
|
Successfully created backport PR for |
My PR fixed the flakiness issue, so it's time to be strict tree-sitter/tree-sitter#5174
My PR fixed the flakiness issue, so it's time to be strict tree-sitter/tree-sitter#5174
Hello,
The test suite of
tsdlruns many builds of the same grammars (same name ofc) in parallel, and the tests were flaky, which required me to put a--retries 2defensively, and it's not always guaranteed to work.I skimmed through the cli, and I figured there shouldn't be a reason to disallow concurrent builds like this.
So I did the fix, and it works for me now.
The choices I made with the hashing and lock file names are ofc arbitrary :)