"SELECT 1;;" means (rc *SQLiteRows) Next() enters an infinite loop (SQLITE_MISUSE)

[otoolep]

Spotted by rqlite/rqlite#1250

Apply the diff below, execute time go test -run TestQueryer and it will never return. Without the diff the test it runs fine, and exits normally. From instrumenting my own code (rqlite) it seems rows.Next() never returns an error. Should the code be more robust against this, or am I doing something wrong?

$ git diff
diff --git a/sqlite3_test.go b/sqlite3_test.go
index 326361e..df317f2 100644
--- a/sqlite3_test.go
+++ b/sqlite3_test.go
@@ -1114,9 +1114,7 @@ func TestQueryer(t *testing.T) {
        if err != nil {
                t.Error("Failed to call db.Exec:", err)
        }
-       rows, err := db.Query(`
-               select id from foo order by id;
-       `)
+       rows, err := db.Query(`SELECT 1;;`)
        if err != nil {
                t.Error("Failed to call db.Query:", err)
        }

[otoolep]

Digging into this I see rv := C._sqlite3_step_internal(rc.s.s) return (21) SQLITE_MISUSE which sends the code down the if rv != C.SQLITE_ROW branch. Then it calls rv = C.sqlite3_reset(rc.s.s) which returns (0) SQLITE_OK. And then nil is returned by Next(). Then Next() is called again, and we keep going forever.

According to the SQLite docs, getting (21) SQLITE_MISUSE is bad.

If SQLite ever returns SQLITE_MISUSE from any interface, that means that the application is incorrectly coded and needs to be fixed. Do not ship an application that sometimes returns SQLITE_MISUSE from a standard SQLite interface because that application contains potentially serious bugs.

Does db.Query("SELECT 1;;") violate the API for go-sqlite3? If so, how?

[rittneje]

This is the same underlying issue as discussed in #950 and #1133. Namely, this library is not using sqlite3_prepare_v2 correctly.

That said, passing multiple statements to Query is not something I would recommend, as the behavior is ill-defined.

[otoolep]

Dupe of #950

[otoolep]

OK, thanks. So is there a reason #950 isn't fixed? You seem to have identified the issue, but it's not clear if it's much more difficult to fix that it seems.

[otoolep]

I can patch my copy of my fork, but it's just rather unfortunate that a simple string can bring down such an important library as this one.

[rittneje]

Because I have not had the time to deal with it. Feel free to raise a PR and I'll take a look.

[otoolep]

OK, let me see what I can do.

[danlzh]

Currently is there a way to detect this? Encountered the same issue when executing malformed empty SQL