Create the action only if the text change creation is successful. #22343
Conversation
| @@ -5017,6 +5017,59 @@ namespace ts.projectSystem { | |||
| service.openExternalProject({ projectFileName: "p", rootFiles: [toExternalFile(f1.path)], options: { importHelpers: true } }); | |||
| service.checkNumberOfProjects({ externalProjects: 1 }); | |||
| }); | |||
|
|
|||
| it("should not crash when getting code fix from dynamic import", () => { | |||
There was a problem hiding this comment.
This issue didn't have anything to do with the server, just a bug in a codefix, so would be better as a fourslash test. See the many existing codeFixInferFromUsage... fourslash tests.
There was a problem hiding this comment.
fourslash is not easily equipped right now to test the negative scenarios of the codeFix. The current error happens because one of the action's change is undefined and we try to map the change to code edit. I am not sure how i test that from fourslash with current infrastructure.
There was a problem hiding this comment.
I got a failing fourslash test in the master branch:
/// <reference path='fourslash.ts' />
// @noImplicitAny: true
// @Filename: /a.ts
////export class D {}
////export default new D();
// @Filename: /b.ts
////export class C {
//// [|set x(val) {}|]
//// method() { this.x = import("./a"); }
////}
goTo.file("/b.ts");
verify.not.codeFixAvailable();codeFixInferFromUsageInaccessibleTypes.ts appears to be testing a similar thing, so you might add this onto that one.
Make change for the infer type from usage could return undefined even if type is present if the type cannot be named Fixes #22184
sheetalkamat
force-pushed
the
codeFixActionNotUndefined
branch
from
3049410 to
7094442
Compare
Make change for the infer type from usage could return undefined even if type is present if the type cannot be named
Fixes #22184