https://bugs.openldap.org/show_bug.cgi?id=9949
Issue ID: 9949
Summary: MDB_RDONLY txn segfaults on newly created database
Product: LMDB
Version: unspecified
Hardware: All
OS: All
Status: UNCONFIRMED
Keywords: needs_review
Severity: normal
Priority: ---
Component: liblmdb
Assignee: b...@openldap.org
Reporter: jeffrey.reyno...@ticketmaster.com
Target Milestone: ---
The very simple code will cause a seg fault.
```
auto env = create_env("env_name");
// creates the environment. not included here because this part is in rust
// it will open or create the database. i don't think the problem lies in
here.
MDB_txn* txn{};
mdb_txn_begin(*env, nullptr, MDB_RDONLY, &txn);
MDB_dbi dbi{};
mdb_dbi_open(txn, "db_name", MDB_CREATE, &dbi);
```
This segfaults on `liblmdb/mdb.c:11050`. Specifically `tracked->mc_next = *tp;`
However, the problem isn't in mdb_dbi_open, it is failing because mt_cursors is
never initialized.
A small change ` mdb_txn_begin(*env, nullptr, 0, &txn);` and mt_cursors will
be initialized with the default env->me_txn0, that has a properly initialized
mt_cursors, per this line `liblmdb/mdb.c:5581`, `txn->mt_cursors = (MDB_cursor
**)(txn->mt_dbs + env->me_maxdbs);`
for the MDB_RDONLY transaction, it looks like it will initialize mt_cursors
_if_ it happens to have a parent, `liblmdb/mdb.c:3178`, but otherwise it leaves
it uninitialized.
Is this a bug, or do have i have to a parent to start a readonly transaction on
a new database?
--
You are receiving this mail because:
You are on the CC list for the issue.
