Fix checker initialization crash #46973
Conversation
typescript-bot
added
the
For Milestone Bug
There was a problem hiding this comment.
The work we do when producing diagnostics sometimes has side effects, like pushing things into caches, which can affect other calculations down the line, so I'm usually wary against produceDiagnostics checks that aren't immediately around an error call, as it can the a source of reeeeeaaally hard to track bugs where hover info and error info on a symbol don't match (hover data comes from a non-diagnostics producing checkers, while errors do). Specifically here, there's a nested conditional checkResolvedBlockScopedVariable that might toss something into a cache which now we'd avoid doing - it's probably fine until proven otherwise, since everything in the guarded branch is supposed to just be to generate an error message, I'd just keep it in mind in case something weird crops up in the future.
|
@typescript-bot cherry-pick this to release-4.5 |
|
Heya @andrewbranch, I've started to run the task to cherry-pick this into |
|
Hey @andrewbranch, I've opened #47005 for you. |
| excludeGlobals = false): Symbol | undefined { | ||
| return resolveNameHelper(location, name, meaning, nameNotFoundMessage, nameArg, isUse, excludeGlobals, getSymbol); | ||
| excludeGlobals = false, | ||
| getSpellingSuggstions = true): Symbol | undefined { |
There was a problem hiding this comment.
I'm obviously very late to this PR but is this argument ironically named? 😛
* Fix checker initialization crash * Move checks to a place that makes more sense
Fixes #46587
The auto import provider was trying to generate spelling suggestions for
IArgumentssince it runs withnoLib: true. Not sure why this happens only sometimes, but (a) there’s no point in trying to come up with spelling suggestions for global symbols, and (b) there’s no point in trying to come up with spelling suggestions for anything inside a non-diagnostics-producing checker.