@agorrod, I had a question about __wt_clsm_open I didn't want to lose, so I'm creating an issue.
Obviously, not urgent.
In __wt_clsm_open, there's this code:
/*
* The tree's dsk_gen starts at one, so starting the cursor on zero
* will force a call into open_cursors on the first operation.
*/
clsm->dsk_gen = 0;
WT_STATIC_ASSERT(offsetof(WT_CURSOR_LSM, iface) == 0);
WT_ERR(__wt_cursor_init(cursor, cursor->uri, owner, cfg, cursorp));
if (0) {
err: __wt_lsm_tree_release(session, lsm_tree);
if (clsm != NULL) {
clsm->lsm_tree = NULL;
WT_TRET(__clsm_close(cursor));
}
}
This is probably wrong, I think we should be able to call free(clsm) instead of *clsm_close; if I'm wrong, we shouldn't call *wt_lsm_tree_release, as _clsm_close does that, and in this case I think the current code would result in _wt_lsm_tree_release being called twice.