Howard Chu authored 10 years ago and Quanah Gibson-Mount committed 10 years ago

Aliases pointing within the original search scope could cause
the same entry to be returned multiple times. Check for subtree
scope and duplicate scopes.

Howard Chu authored 10 years ago and Quanah Gibson-Mount committed 10 years ago

Must save/restore mcd cursor if we're doing a scope-based search.

    Add sc_writewait callback

    Invoked before a blocked writer waits for socket to be writable.
    Use in back-mdb to release reader txn while waiting.

Howard Chu authored 10 years ago and Quanah Gibson-Mount committed 10 years ago

mdb_idscopes was remembering the IDs of every entry it checked;
it should only have been saving the IDs of entries that were
actually in the search scope. The extra entries filled the scopes
array, causing a loop searching for a parent entry that was
never inserted.

Howard Chu authored 11 years ago and Quanah Gibson-Mount committed 11 years ago

If multiple ops used the same opinfo the refcount wasn't being decremented

Howard Chu authored 12 years ago and Quanah Gibson-Mount committed 12 years ago

uninit'd size in dn2id_walk, bogus pagedresults

Howard Chu authored 12 years ago and Quanah Gibson-Mount committed 12 years ago

Walk subtree if number of subtree entries is smaller than
number of index candidates.

Howard Chu authored 12 years ago and Quanah Gibson-Mount committed 12 years ago

This is only a 10% gain, but still worth having for now.

Howard Chu authored 12 years ago and Quanah Gibson-Mount committed 12 years ago

Fix search_aliases invocation

Howard Chu authored 13 years ago and Quanah Gibson-Mount committed 13 years ago

Use malloc'd memory, stored persistently in thread-local storage.

Howard Chu authored 13 years ago and Quanah Gibson-Mount committed 13 years ago

Tighten up entry_alloc/entry_decode
Track parent nodes in idscopes

Howard Chu authored 13 years ago and Quanah Gibson-Mount committed 13 years ago

Instead of iterating thru potentially many nonexistent IDs