ISSUE DESCRIPTION Fatal crash in (*Program).SingleThreaded on snapshot update [code]…
- User
ISSUE DESCRIPTION
Fatal crash in
(*Program).SingleThreadedon snapshot updateinternal/compiler.(*Program).SingleThreaded() internal/compiler/program.go:416 internal/compiler.NewProgram() internal/compiler/program.go:274 internal/project.(*Project).CreateProgram() internal/project/project.go:376 internal/project.(*ProjectCollectionBuilder).updateProgram.func2.1() internal/project/projectcollectionbuilder.go:1055 internal/project/dirty.(*Box[...]).Change() internal/project/dirty/box.go:41 internal/project.(*ProjectCollectionBuilder).updateProgram.func2() internal/project/projectcollectionbuilder.go:1052 internal/project/dirty.(*Box[...]).Locked() internal/project/dirty/box.go:57 internal/project.(*ProjectCollectionBuilder).updateProgram() internal/project/projectcollectionbuilder.go:1051 internal/project.(*ProjectCollectionBuilder).DidRequestFile() internal/project/projectcollectionbuilder.go:397 internal/project.(*Snapshot).Clone() internal/project/snapshot.go:353 internal/project.(*Session).updateSnapshot() internal/project/session.go:1101 internal/project.(*Session).getSnapshot() internal/project/session.go:819 internal/project.(*Session).getSnapshotAndDefaultProject() internal/project/session.go:877 internal/project.(*Session).GetLanguageService() internal/project/session.go:890 internal/lsp.init.func1.registerLanguageServiceDocumentRequestHandler[...].13() internal/lsp/server.go:826 internal/lsp.(*Server).handleRequestOrNotification() internal/lsp/server.go:669Interface
No new interfaces are introduced.
ADDITIONAL ADVICE
Since you are given a git repository, you can use git commands to simplify your work. For example, if you made a mistake and want to revert your changes, you can use
git checkout <file>to restore the file to its original state. You can also reset all changes in the repository usinggit reset --hardcommand. Additionally, you can usegit stashandgit stash popto temporarily save your changes and restore them later. Finally, you can see the changes that you've made (which will be submitted when you callsubmit) usinggit statusorgit diffcommands. However you don't need to usegit addorgit commitbefore submitting your changes. In fact,submitonly submits currently unstaged changes because it usesgit diffto get the changes that need to be submitted.CHECKLIST
Before submitting your solution, please go over the following checklist and make sure you've done everything:
- If an error was reported in the issue description, I have successfully reproduced it.
- If an error was reported in the issue description, I have confirmed that my fix resolves the error.
- I have thought about edge cases and made sure my fix handles them as well.
- I have run existing tests in the repository that might have been affected by the change I made and confirmed that they pass. I want you to list every bullet from this checklist and write your reasoning for why you think you did it or didn't need to.
Repository has been uploaded and your shell is currently at the repository root. Time to solve the issue!
(Current directory: /typescript-go, current file: none) bash-$
Assistant
Now let me write a temporary reproduction test.
Assistant
Now update Clone and the getter.
Assistant
Summary
Root cause.
Project.getCommandLineWithTypingsFiles()lazily computed the "command line augmented with ATA typings files" behind async.Once, while two places inProjectCollectionBuilderinvalidate that cache by setting the field tonil(updateInferredProjectRootsandupdateProgramon config change).Since
dirty.Box.Change/SyncMapEntry.changeLockedonly clone theProjectthe first time it becomes dirty, a single snapshot update can:- create a program for the (already dirty) project → the
Oncefires and the augmented command line is cached, - later invalidate it in the same pass (
p.commandLineWithTypingsFiles = nil) on the very same instance, - create a program again →
Once.Dono longer runs, sogetCommandLineWithTypingsFiles()returns nil →compiler.NewProgramgetsConfig == nil→p.Options()is nil → nil-pointer deref in(*Program).SingleThreaded— exactly the reported crash stack.
(The same
Oncealso made ATA typing updates silently stale, sinceDidUpdateATAStatechangestypingsFileswithout touching the cache.)Fix (
internal/project/project.go): replaced the one-shotsync.Oncewith a mutex-protected cache that is keyed on its inputs (CommandLineidentity andtypingsFiles), so it recomputes as many times as needed and can never return a stale/nil value. Cache keys are carried throughClone.Verified with a temporary reproduction (open a JS file whose
package.jsontriggers ATA typings, edit it, then request the open file plus a closed file in one snapshot update): it panicked with the exact stack from the issue before the change and passes afterwards, with the program correctly containing both the typings file and the newly added root.Checklist
- If an error was reported in the issue description, I have successfully reproduced it. Yes — I built a temporary test in
internal/projectdriving the realSessionAPI which produced the identical panic (SingleThreaded→NewProgram→CreateProgram→dirty.(*Box).Change→updateProgram→DidRequestFile). - If an error was reported in the issue description, I have confirmed that my fix resolves the error. Yes — the same scenario passes after the change, and the resulting program contains the expected files (typings + new root), so the fix is correct, not just crash-suppressing.
- I have thought about edge cases and made sure my fix handles them as well. The cache is now keyed on both inputs, so it also handles:
CommandLinereplacement (configured-project config reload, inferred-root changes),typingsFileschanges from ATA (previously silently stale), repeated invalidation within one snapshot pass, and cloned projects inheriting the cache. Access is mutex-protected so concurrent program creation cannot produce two different command-line pointers (which would defeat program reuse);go vet(copylocks) and-raceruns are clean. The existing explicit= nilinvalidations remain valid and harmless. - I have run existing tests in the repository that might have been affected and confirmed that they pass. Yes —
go test ./internal/project/...(also with-race),./internal/lsp/...,./internal/ls/...,./internal/fourslash/...all pass;go build ./...andgofmtare clean. No existing tests were modified and my temporary test file was removed.
- create a program for the (already dirty) project → the
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.
Identical first prompt
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.5-35B-A3BCompare side by side →
Qwen/Qwen3.5-35B-A3BCompare side by side →