Bun crashes opening DuckDB

[fleitz]

How can we reproduce the crash?

Increase the number of microtasks until it crashes. Usually works fine below 5, definitely crashes at 100 or more.

import { DuckDBInstance } from '@duckdb/node-api'
import { randomUUIDv7 } from "bun";

const run = async () => {
    for (let i = 0; i < 10000; i++) {
        const id = randomUUIDv7();
        const path = `./${id}.db`
        const db = await DuckDBInstance.create(path)
        const connection = await db.connect()
        await connection.close()
    }
}

for(let i = 0; i < 10; i++) {
    queueMicrotask(() => {
        run()
    })
}

gdb --args bun run index.ts

__memmove_evex_unaligned_erms () at ../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:872
872     ../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S: No such file or directory.
(gdb) backtrace
#0  __memmove_evex_unaligned_erms () at ../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:872
#1  0x00007fff9c39b8d0 in std::string::_Rep::_M_clone(std::allocator<char> const&, unsigned long) () from /lib/x86_64-linux-gnu/libstdc++.so.6
#2  0x00007fff9c39c344 in std::string::assign(std::string const&) () from /lib/x86_64-linux-gnu/libstdc++.so.6
#3  0x00007fff9dc73d12 in duckdb::DatabaseInstance::Configure(duckdb::DBConfig&, char const*) ()
   from /home/fred/src/duckdb_crash/node_modules/@duckdb/node-bindings-linux-x64/libduckdb.so
#4  0x00007fff9dc7dbef in duckdb::DatabaseInstance::Initialize(char const*, duckdb::DBConfig*) ()
   from /home/fred/src/duckdb_crash/node_modules/@duckdb/node-bindings-linux-x64/libduckdb.so
#5  0x00007fff9dc7e707 in duckdb::DuckDB::DuckDB(char const*, duckdb::DBConfig*) ()
   from /home/fred/src/duckdb_crash/node_modules/@duckdb/node-bindings-linux-x64/libduckdb.so
#6  0x00007fff9dbaafc6 in duckdb_open_internal(duckdb::DBInstanceCacheWrapper*, char const*, _duckdb_database**, _duckdb_config*, char**) ()
   from /home/fred/src/duckdb_crash/node_modules/@duckdb/node-bindings-linux-x64/libduckdb.so
#7  0x00007fffa802a741 in OpenWorker::Execute() () from /home/fred/src/duckdb_crash/node_modules/@duckdb/node-bindings-linux-x64/duckdb.node
#8  0x00007fffa802a8c5 in Napi::AsyncWorker::OnAsyncWorkExecute(napi_env__*, void*) ()
   from /home/fred/src/duckdb_crash/node_modules/@duckdb/node-bindings-linux-x64/duckdb.node
#9  0x000000000308aff9 in ?? ()
#10 0x0000000000000000 in ?? ()

Relevant log output

Stack Trace (bun.report)

Bun v1.2.2 (c1708ea) on linux x86_64 [RunCommand]

Segmentation fault at address 0x081DA570

• Signals.cpp:0: WTF::jscSignalHandler

Features: tsconfig_paths, process_dlopen, Bun.stderr, Bun.stdout, fetch, http_server, jsc, transpiler_cache, tsconfig, tsconfig

_{Sentry Issue: BUN-7G}

[jraymakers]

I don't know why the crash is occurring, but the code above is an anti-pattern. It's recommended to create a single DuckDB instance per process, and use the ATTACH command to load (possibly multiple) databases into that instance. Creating thousands of DuckDB instances, especially at such a high rate, is not advised.

[RiskyMH]

Probably the same issue of #13910

Also if it works in Node but not Bun, please only open issues here :)