* chore: map mobile browser data to their desktop version
* chore: skip android until upstream support is resolved
* Revert "chore: skip android until upstream support is resolved"
This reverts commit 436e9af8297b284a426b5a3240e2a550a20fca86.
* Update fixtures
Co-authored-by: Nicolò Ribaudo <nicolo.ribaudo@gmail.com>
* fix(babel-parser): avoid state.clone() to clone the whole token store
Fixed the performance issue on large input when turned on option {tokens: true} and typescript plugin which uses quite a few state.clone().
* test(babel-parser): turn on 2 typescript tests with tokens:true
The output.json is generated by old master to make sure no regression.
* fix(babel-parser): avoid duplicated tokens trapped by mainly typescript/flow plugins
* test(babel-parser): update output.json to latest master result
* chore(babel-parser): improve performance by storing tokensLength in state
* Bug replication test
* Simplify test
* Fixes#10896
* Merge path.crawl `ClassDeclaration` and `Declaration` visitors
Merged to avoid subtle assumption that `Declaration` is called before `ClassDeclaration`
* Move registartion of class id in crawl to BlockScoped
* Add some assertions to crawl test
Fixes issue 10989 where the only the identifier in a member expression that is the value of an object property would be parsed. Removing checkExpressionErrors in parseExprSubscripts results in the subscript also being parsed.
* Fix rest parameters indexing with TypeScript 'this parameter'
The TypeScript [this parameter][0] is a fake parameter and should not be
taken into account when counting the function parameters. This patch
skips it by using the condition taken from the `transform-typescript`
plugin.
Note: since the `transform-typescript` plugin is removing this kind of
parameter, including it before the `transform-parameters` plugin solves
the issue. This patch fixes the issue in other cases.
[0]: https://www.typescriptlang.org/docs/handbook/functions.html#this-parameters
* nit: improve the 'this parameter' detection condition
* improve performance by checking the parameter list length before
accessing it
* simplify the test a bit by using the `isIdentifier` second
parameter
I really don't like this commit, but import() is currently breaking
our publish script.
When "normal" tests we are transpiling import() so that it works
with Jest. We can't do it while publishing because we need to
publish the untranspiled import() so that it can load real .mjs files.
Follow up to #10903
* Added precedence for nullish-coalescing-operator
* Made precedence equal to logical OR
* Renamed the folders made for testing
* Fixed the output test file of nullish-coalescing op
* refactor: replace refShorthandDefaultPos by refExpressionErrors
* fix: duplicate __proto__ keys should be allowed in patterns
* docs: add comments for ExpressionErrors.doubleProto [ci-skip]
* test: add more test for coverage
