Disallow duplicated AST nodes #11807
Conversation
JLHwung
added
the
PR: Bug Fix 🐛
|
This pull request is automatically built and testable in CodeSandbox. To see build info of the built libraries, click here or the icon next to each commit SHA. Latest deployment of this branch, based on commit 1e796b3:
|
|
Build successful! You can test your changes in the REPL here: https://babeljs.io/repl/build/25897/ |
| for (const bareSuper of bareSupers) { | ||
| bareSuper.insertAfter(nodes); | ||
| if (isFirst) { |
There was a problem hiding this comment.
The test legacy-class-prototype-properties/child-classes-properties is failing if we change these to bareSuper.insertAfter(nodes.map(n => t.cloneNode(n))) unconditionally. The current approach happens to make our test cases happy and also slightly faster because we get rid of redundant clone.
However, I think it reveals an issue which is not caught in our current test cases.
I suggest we address that in a separate PR or even let it stay in the backlog since it is related to legacy decorators only.
I am not familiar with decorators part, suggestions are definitely welcome.
There was a problem hiding this comment.
It is probably failing because in the decorators plugin we use a WARNING_CALLS WeakSet to track warnings: if we clone the node we cannot find it anymore in the set.
| for (const bareSuper of bareSupers) { | ||
| bareSuper.insertAfter(nodes); | ||
| if (isFirst) { |
There was a problem hiding this comment.
It is probably failing because in the decorators plugin we use a WARNING_CALLS WeakSet to track warnings: if we clone the node we cannot find it anymore in the set.
JLHwung
force-pushed
the
duplicated-nodes
branch
from
78ff568
to
1e796b3
Compare
github-actions
bot
added
the
outdated
As a follow-up to #7149, this PR enables duplicated node checks against all testcases and then fixes all duplicate AST errors.
I didn't add a regression test of #11506 because it is covered in this test case. I realized that the duplicated node checks are not effective when I am investigating #11506, which should have been captured by our test runner.