False errors due to off-by-one-character in identifiers or types

[2-5]

Volar 0.40.5

I get frequent false errors due to Volar complaining about invalid identifiers due to off-by-one-character errors, like in the screenshot:

The fix is to restart VS Code, closing and reopening the source file with the reported false error doesn't work due to some form of cache.

Other example that happend: complaining about unknown type nboolean instead of boolean which is in source code.

[johnsoncodehk]

Since v0.40.2 we have handle incremental update logic in language server for support incremental parser and resolve diagnosis display flicker (#1718), it should be the reason. Please let me know if you can reproduce this problem stably and how.

After v0.40.6 you can setting "volar.vueserver.textDocumentSync": "full" to disable incremental update.

[Sun-ZhenXing]

Same problem. In recent updates, this has become more common. I hope it will be solved soon.
Restarting VS Code has been difficult to fix and I keep getting a lot of errors in the <template> element that actually compile fine.

[johnsoncodehk]

@Sun-ZhenXing can you reproduce this problem in https://github.com/johnsoncodehk/volar-starter?

[johnsoncodehk]

@Sun-ZhenXing I think this is a different problem and it should be fixed in v0.40.8.

[johnsoncodehk]

@2-5 Since I haven't seen other similar reports, I'm guess it may no longer occur, can you try revert volar.vueserver.textDocumentSync setting to and see if this problem reproduce?

[rchl]

I did have it happen in 1.0.7 yesterday in a *.ts file.

[johnsoncodehk]

@rchl It would be very helpful if you can provide the LSP logs that record at the moment the problem occurred.

[2-5]

A few days later I did revert the textDocumentSync setting, to see if the issue was fixed by the new versions of Volar. It seems so, I haven't seen this particular issue since, but at the same time I didn't do any heavy refactorings either (lots of edits in one file) which appeared to have triggered it.

[rchl]

@rchl It would be very helpful if you can provide the LSP logs that record at the moment the problem occurred.

I'm sure it would but I can't reproduce it on demand and once I notice it, I might not have logs reaching far enough for them to be useful.