`/dump` on Turso Cloud writes the first FTS5 shadow table twice, and still emits unescaped virtual-table SQL

Open
#2,287 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 11 days

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
42/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Active
Tech stack
c, rust

Research direction

Start by comparing the Cloud /dump output with the clean turso dev result, then inspect change 63ba6b8d5 and the /dump implementation. Done means the virtual-table SQL is escaped and the first FTS5 shadow-table block is emitted only once, with a regression test or reproducible check covering restoration with sqlite3.

Written by the indexing model from the issue text.

Description

Summary

A dump of a libSQL database on Turso Cloud that contains an FTS5 table cannot be restored with sqlite3 without errors. Two things are wrong in the output:

  1. The block for the first shadow table (<name>_data: its CREATE TABLE IF NOT EXISTS and all of its INSERT rows) is written twice, back to back and byte-identical. Restoring fails with UNIQUE constraint failed: <name>_data.id.
  2. The virtual-table definition is written without escaping its single quotes. This was fixed in 63ba6b8d5 ("fix: pass argument through Escaped interface", 2025-10-08), but no libsql-server release contains it (the latest is v0.24.32 from 2025-02-14) and Turso Cloud still produces the old output as of 2026-10-04.
Details

The table was created like this, so the virtual table precedes its shadow tables in sqlite_schema (rowid order: notes_fts, notes_fts_data, notes_fts_idx, notes_fts_docsize, notes_fts_config):

CREATE VIRTUAL TABLE notes_fts USING fts5(
  title,
  summary,
  content='notes',
  content_rowid='rowid',
  tokenize='trigram'
);

Dump taken with turso db shell "https://<db>.turso.io?authToken=…" .dump (CLI v1.0.33, which prints the response of GET /dump as is), region aws-ap-northeast-1. Table names changed, structure unchanged:

PRAGMA writable_schema=ON;
INSERT INTO sqlite_schema(type,name,tbl_name,rootpage,sql) VALUES('table','notes_fts','notes_fts',0,'CREATE VIRTUAL TABLE notes_fts USING fts5(
  title,
  summary,
  content='notes',
  content_rowid='rowid',
  tokenize='trigram'
)');
CREATE TABLE IF NOT EXISTS 'notes_fts_data'(id INTEGER PRIMARY KEY, block BLOB);
INSERT INTO notes_fts_data VALUES(1,X'…');
… 30 rows
CREATE TABLE IF NOT EXISTS 'notes_fts_data'(id INTEGER PRIMARY KEY, block BLOB);
INSERT INTO notes_fts_data VALUES(1,X'…');
… the same 30 rows
CREATE TABLE IF NOT EXISTS 'notes_fts_idx'(segid, term, pgno, PRIMARY KEY(segid, term)) WITHOUT ROWID;
…
  • Seen in two dumps taken a day apart (2026-10-03 and 2026-10-04). No other table is repeated. A second Cloud database without virtual tables dumps cleanly.
  • Not reproducible locally: the same table created in the same order and served by turso dev (sqld 0.24.32) is dumped once.
Effect

sqlite3 -bail restored.db < dump.sql stops at the virtual-table statement with a syntax error and, once that statement is fixed by hand, at the repeated rows. Without -bail the restore finishes, but the FTS table is missing while its triggers remain, so every insert into the content table fails with no such table: main.notes_fts.

Ask
  • A libsql-server release, and a Turso Cloud deploy, that include 63ba6b8d5.
  • A look at why the Cloud /dump repeats the first shadow table.
Dominant language
C
Stars
17.3k
Forks
562
Avg merge
10d 15h
Merged PRs (30d)
1

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from tursodatabase/libsql

All issues in tursodatabase/libsql

Similar issues

More C issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.