Another "FATAL ERROR: v8::ToLocalChecked Empty MaybeLocal" error

[achou11]

Looks like I'm running into a similar error. Apologies for potential redundancy (seem to be quite a few issues with this error message but with different causes)

Node version: 12.18.3
better-sqlite3 version: 7.5.3
OS: MacOS 11.6.7 (Big Sur), Intel

FATAL ERROR: v8::ToLocalChecked Empty MaybeLocal.
 1: 0x1011c2ff5 node::Abort() (.cold.1) [/Users/andrewchou/Library/Caches/fnm_multishells/17210_1657814232551/bin/node]
 2: 0x10009fbc9 node::Abort() [/Users/andrewchou/Library/Caches/fnm_multishells/17210_1657814232551/bin/node]
 3: 0x10009fd2f node::OnFatalError(char const*, char const*) [/Users/andrewchou/Library/Caches/fnm_multishells/17210_1657814232551/bin/node]
 4: 0x1001e60e0 v8::V8::ToLocalEmpty() [/Users/andrewchou/Library/Caches/fnm_multishells/17210_1657814232551/bin/node]
 5: 0x106807e46 Database::ThrowSqliteError(Addon*, char const*, int) [/Users/andrewchou/GitHub/digidem/mapeo-map-server/node_modules/better-sqlite3/build/Release/better_sqlite3.node]
 6: 0x1068092a1 Statement::JS_run(v8::FunctionCallbackInfo<v8::Value> const&) [/Users/andrewchou/GitHub/digidem/mapeo-map-server/node_modules/better-sqlite3/build/Release/better_sqlite3.node]
 7: 0x10024f438 v8::internal::FunctionCallbackArguments::Call(v8::internal::CallHandlerInfo) [/Users/andrewchou/Library/Caches/fnm_multishells/17210_1657814232551/bin/node]
 8: 0x10024e9f9 v8::internal::MaybeHandle<v8::internal::Object> v8::internal::(anonymous namespace)::HandleApiCallHelper<false>(v8::internal::Isolate*, v8::internal::Handle<v8::internal::HeapObject>, v8::internal::Handle<v8::internal::HeapObject>, v8::internal::Handle<v8::internal::FunctionTemplateInfo>, v8::internal::Handle<v8::internal::Object>, v8::internal::BuiltinArguments) [/Users/andrewchou/Library/Caches/fnm_multishells/17210_1657814232551/bin/node]
 9: 0x10024e162 v8::internal::Builtin_Impl_HandleApiCall(v8::internal::BuiltinArguments, v8::internal::Isolate*) [/Users/andrewchou/Library/Caches/fnm_multishells/17210_1657814232551/bin/node]
10: 0x1009d3ab9 Builtins_CEntry_Return1_DontSaveFPRegs_ArgvOnStack_BuiltinExit [/Users/andrewchou/Library/Caches/fnm_multishells/17210_1657814232551/bin/node]
11: 0x100959364 Builtins_InterpreterEntryTrampoline [/Users/andrewchou/Library/Caches/fnm_multishells/17210_1657814232551/bin/node]
12: 0x100959364 Builtins_InterpreterEntryTrampoline [/Users/andrewchou/Library/Caches/fnm_multishells/17210_1657814232551/bin/node]
13: 0x100959364 Builtins_InterpreterEntryTrampoline [/Users/andrewchou/Library/Caches/fnm_multishells/17210_1657814232551/bin/node]

Haven't figured out a minimal reproduction but can make note of the thing that seems to cause it. I have a set of 4 statements that are run within a transaction (better-sqlite3 transaction, not hardcoded). There are other statements in this transaction, but this set seems to be causing issues. The nature of the statements are as follows:

1. Create a view. This does a join on a subquery and uses some basic json_each stuff
2. Delete some records using view from (1) to do a subquery for the where clause
3. Similar to (2) but deleting from a different table
4. Drop view

Other important notes:

• WAL mode is enabled
• Seeing this failure when in the context of tests that I'm running (using tape + ts-node). sometimes the test works and i don't see this error. but i'd say it comes up once every three or four attempts

Entirely possible that this is a red herring. Also, could be approaching this in a non-idiomatic way and maybe there's a better way of doing the above that avoids the issue. Maybe using a CTE might be a better approach, although it seemed to be much slower than using a view since it has to be run for each deletion statement.

Apologies for not having a minimum repro available, but will try for a bit and see if I can create one

Originally posted by @achou11 in #281 (comment)

[mceachen]

FWIW Node 12 is EOL (and has for several months, and several security issues have not been back-patched to that series). Can you reproduce on a current Node and the latest version of this project?

[achou11]

FWIW Node 12 is EOL (and has for several months, and several security issues have not been back-patched to that series

Yeah I'm aware of that - unfortunately it's a constraint of the project for the time being 🥲 don't think i specifically tested it locally but I think our CI (which runs this test on Node 12, 14, and 16) encountered the same issue.

will try to upgrade the module and see if that helps. stay tuned...

[JoshuaWise]

The stack trace is helpful, but unfortunately I can't diagnose the problem much without a reproducible example. The code path that's failing is working perfectly in all our tests.

[achou11]

I'll try to spend some time attempting to repro it, although I'm traveling for a bit soon so may be a while 😅

[achou11]

Not a minimal repro but just for me to look back on - here's a snippet from what i'm working with that currently fails (not all the time, but maybe ~50%):

  const deleteStyleTransaction = db.transaction(() => {
     // Some stuff before this...

    db.prepare(
      `
      CREATE VIEW DeletableTilesetIds AS
      SELECT SourceToTilesetIdOuter.value AS tilesetId, Style.id AS styleId
      FROM Style, json_each(Style.sourceIdToTilesetId, '$') AS SourceToTilesetIdOuter
      JOIN (
        SELECT SourceToTilesetIdInner.value AS tilesetId, COUNT(SourceToTilesetIdInner.value) AS freq
        FROM Style, json_each(Style.sourceIdToTilesetId, '$') AS SourceToTilesetIdInner
        GROUP BY SourceToTilesetIdInner.value
      ) AS TilesetIdFreq ON SourceToTilesetIdOuter.value = TilesetIdFreq.tilesetId
      WHERE freq = 1;
    `
    ).run()
    db.prepare(
      'DELETE FROM Tile WHERE tilesetId ' +
        'IN (SELECT tilesetId FROM DeletableTilesetIds WHERE styleId = ?)'
    ).run(id)
    /**
     * Deleting this specific statement seems to prevent the issue from happening 🤔
     */
    db.prepare(
      'DELETE FROM Tileset WHERE id ' +
        'IN (SELECT tilesetId FROM DeletableTilesetIds WHERE styleId = ?)'
    ).run(id)
    db.prepare('DROP VIEW DeletableTilesetIds').run()

    // Some stuff after this...
  })

  deleteStyleTransaction()