Confusing error message for labels used before definition

[paulvanbrenk]

The following gives me an unexpected "TS1007: Jump target cannot cross function boundary error" in TS 3.3.33333

function foo() {
    for (let i = 0; i < 10; i++) {
        console.log(`${i}`);
        continue loopend;
    }

    loopend:
    console.log('end of loop');
}

Further investigation... you can jump to the start of the loop, so the error I would argue is that the error message is confusing.

shareable link

[DanielRosenwasser]

Did you try adding 3s to the version to see if the error gets better?

Seems like a reasonable request.

[DanielRosenwasser]

Here's a better message for this specific case, where the containing function of the label reference is the same as that of the labeled statement:

A label cannot be referenced prior to its declared location.

To find if it's applicable, walk up each scope until the first functionlike (isFunctionLike) or classlike (isClassLike) declaration and look for LabelledStatements in that scope.

We can also provide a related span. We already have a diagnostic for

{0} is declared here.

[paulvanbrenk]

?????? Get Outlook for Android<https://aka.ms/ghei36>

…

________________________________ From: Daniel Rosenwasser <[email protected]> Sent: Thursday, March 14, 2019 6:41:32 PM To: Microsoft/TypeScript Cc: Paul van Brenk; Author Subject: Re: [Microsoft/TypeScript] Confusing error message for labels used before definition (#30408) Here's a better message for this specific case. A label cannot be referenced prior to its declared location. We also should provide a related span. We already have a diagnostic for {0} is declared here. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#30408 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AFB5dwK9ovsiqwXTAzc2LlmDu5_corNoks5vWtAcgaJpZM4b1HbC>.

[DanielRosenwasser]

Actually, we already have a better message for break or continue error messages.

[paulvanbrenk]

In 3.4? Get Outlook for Android<https://aka.ms/ghei36>

…

________________________________ From: Daniel Rosenwasser <[email protected]> Sent: Thursday, March 14, 2019 8:13:58 PM To: Microsoft/TypeScript Cc: Paul van Brenk; Author Subject: Re: [Microsoft/TypeScript] Confusing error message for labels used before definition (#30408) Actually, we already have a better message for break or continue error messages. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#30408 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AFB5d5sWoEtjlpcsCEz1AA8hP80x1ghKks5vWuXGgaJpZM4b1HbC>.

[laurafiuza]

Hey -- is this still something that should be implemented? Happy to make changes now if no one else is already

[paulvanbrenk]

Yes, please. This is still an issue in @next

[DanielRosenwasser]

It may be a bit more work than I originally anticipated (since labels aren't placed in any sort of scope), but we're taking PRs if you're up for it.

I guess if you can walk up the tree and look for labels in each enclosing scope, then:

• If the label actually contains the source of the break/continue, but it crosses a function boundary, you can continue to give the same error message. The key difference is that you need to keep walking up the tree to search for the label and don't just stop at a function boundary.

• If the label doesn't contain the origin of the break/continue, then use either

  Diagnostics.A_break_statement_can_only_jump_to_a_label_of_an_enclosing_statement
  

  or

  A_continue_statement_can_only_jump_to_a_label_of_an_enclosing_iteration_statement
  

  (these already exist apparently)

• Otherwise, give a message like

  Cannot find label '{0}'.
  

In the first two, give the appropriate related span mentioned above.

[karthikkp]

@laurafiuza Still working on this?

[karthikkp]

@DanielRosenwasser For point 1, we will have to go through the entire set of statements in the source file to check if the label exists as, as soon as we travel up the tree and go beyond the function boundry we get the source file in the example in the first example. To simply put we will have to check the siblings of the function statement. Is my understanding right?

[DanielRosenwasser]

I would avoid trying to walk through every scope. Try going with this initial suggestion (though ensure you don't go past the SourceFile):

To find if it's applicable, walk up each scope until the first functionlike (isFunctionLike) or classlike (isClassLike) declaration and look for LabelledStatements in that scope.

Just scan through the statements at each level until you hit a function-like/class-like - don't try to walk recursively.

[karthikkp]

I would avoid trying to walk through every scope. Try going with this initial suggestion (though ensure you don't go past the SourceFile):

To find if it's applicable, walk up each scope until the first functionlike (isFunctionLike) or classlike (isClassLike) declaration and look for _LabelledStatement_s in that scope.

Just scan through the statements at each level until you hit a function-like/class-like - don't try to walk recursively.

@DanielRosenwasser This is what is already being done. The cross boundary error is returned as soon as the first functionlike statment is found.

[mInzamamMalik]

you could have thrown a better error message, talking about this: https://stackoverflow.com/questions/59859011/break-says-error-jump-target-cannot-cross-function-boundary-typescript/59859151#59859151