mirror of https://github.com/oxen-io/oxen-core.git
4f6f67d151
It seems that the `m_tinfo` can be null, sometimes, when `m_cursors == &m_wcursors` is true, and the upstream Monero code (which is pure macro) doesn't touch the bool in such a case. For some reason this started segfaulting now, only on macos, only on a release build because of the access into `m_tinfo`. The workaround (which is indeed a correct fix) appears to avoid the segfault, but the segfault could retrigger if that invariant doesn't hold (and it isn't immediately obvious why that invariant *should* hold). This, like pretty much all of the LMDB code, is garbage. |
||
|---|---|---|
| .. | ||
| lmdb | ||
| CMakeLists.txt | ||
| blockchain_db.cpp | ||
| blockchain_db.h | ||
| locked_txn.h | ||
| testdb.h | ||