SIGSEGV (exit 134) in bundled `rightstr()` when the length argument is a large negative number

Open
#1,508 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

@sarthakdev143-lite is already working on this.

Since Nov 14, 2026.

  • #1509 by @sarthakdev143-lite — open

Assessment

Difficulty
4/5
Estimated time
1-2 days
Newbie friendliness
38/100
Issue type
Bug
Clarity
Clearly specified
Activity status
Active
Tech stack
c, java, sqlite
Domain
database

Research direction

Start by reproducing with the one-line SELECT rightstr('abcdef', -1000000000) through the JDBC driver, then find rightstr in the bundled contrib extension-functions.c, focusing on the negative-length branch. The fix belongs in that C code, which the native library build compiles, so you will need to build libsqlitejdbc locally. Done means a large negative length returns an empty string or a clean error with no SIGSEGV, while small negatives still behave as before.

Written by the indexing model from the issue text.

Description

help wanted impacts:JNI

Describe the bug

A single SELECT rightstr(<text>, <large negative>) kills the whole JVM process with SIGSEGV — no Java exception can catch it, no SQLITE_NOMEM is returned, the process just dies (exit code 134).

rightstr is not a core SQLite function; it is supplied by the extension-functions code compiled into the native library shipped in this jar (libsqlitejdbc.dylib). Only the negative length branch of rightstr crashes: small negatives return an empty string, magnitudes ≥ ~2e8 segfault. Positive lengths of the same magnitude and leftstr with the same argument stay clean.

To Reproduce

SELECT rightstr('abcdef', -1000000000);
Connection c = DriverManager.getConnection("jdbc:sqlite::memory:");
try (ResultSet r = c.createStatement()
        .executeQuery("SELECT rightstr('abcdef', -1000000000)")) {
    while (r.next()) { }
}
java -cp "sqlite-jdbc-<version>.jar:." Repro
echo $?   # 134

Expected behavior

Same as small negative arguments: an empty string (rightstr('abcdef', -1) returns length 0). At worst a clean error such as SQLITE_NOMEM / SQLITE_TOOBIG, as happens for positive lengths and for replicate('abcdef', 2147483647).

Logs

JVM crash report (hs_err_pid*.log):

#  SIGSEGV (0xb) at pc=0x0000000104045a74
# Problematic frame:
# C  [sqlite-3.54.1.0-...-libsqlitejdbc.dylib+0x1da74]

Native frames:
C  libsqlitejdbc.dylib+0x1da74
C  libsqlitejdbc.dylib+0x34e70
C  libsqlitejdbc.dylib+0xbebc
J  org.sqlite.core.NativeDB.step(J)I
J  org.sqlite.core.DB.execute(J[Ljava/lang/Object;)I
J  org.sqlite.core.JDBC3Statement.executeQuery

The crash happens inside SQLite expression evaluation (step), not in the JDBC wrapper.

Environment (please complete the following information)

  • OS: macOS 15.6.1 (24G90)
  • CPU architecture: arm64
  • sqlite-jdbc version: 3.54.1.0 (bundled SQLite 3.54.1), JDK 25.0.1+8

Additional context

Stability — one fresh JVM per run, same single statement:

length argument result
-1e6 / -1e7 / -5e7 / -1e8 no crash
-2e8 2/8 crash
-3e8 7/8 crash
-5e8 / -8e8 / -1e9 / -1.5e9 / -2e9 8/8 crash
INT64_MIN (-9223372036854775808) 0/2 crash

Reproduced across driver versions with -1000000000 (3 runs each): 3.49.1.0 → 3/3, 3.51.2.0 → 3/3, 3.53.4.0 → 3/3, 3.54.1.0 → 3/3.

Control group, same magnitude, no crash:

statement result
leftstr('abcdef', -1000000000) 0/5
substr('abcdef', -200000000) 0/3
rightstr('abcdef', 200000000) 0/3
padl('abcdef', 200000000, 'x') 0/3
padr('abcdef', 200000000, 'x') 0/3
padc('abcdef', 2000000000) 0/5
replicate('abcdef', 2147483647) clean SQLITE_NOMEM: a malloc() failed

Since positive lengths of the same size only produce a clean SQLITE_NOMEM/SQLITE_TOOBIG, this is not memory exhaustion: the length/pointer computation on the negative branch goes out of bounds once |N| reaches roughly 2e8.

Suggested fix: handle N < 0 explicitly in rightstr (return an empty string or sqlite3_result_error), and check nByte > db->aLimit[SQLITE_LIMIT_LENGTH] before allocating, as core substr does.

rightstr is not core SQLite — it comes from contrib extension-functions.c, which SQLite itself does not maintain. The crash is inside libsqlitejdbc.dylib, which this project compiles and ships, hence the report here. Found by sqlancer PST mutation testing while probing function parameter domains.

Dominant language
Java
Stars
3.3k
Forks
676
Avg merge
1d 15h
Merged PRs (30d)
18

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 xerial/sqlite-jdbc

All issues in xerial/sqlite-jdbc

Similar issues

More Java issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.