bruns requested changes to this revision.
bruns added inline comments.
This revision now requires changes to proceed.
INLINE COMMENTS
> poboiko wrote in postingdb.cpp:238
> Can it happen that `rc == MDB_NOTFOUND` just after some iterations of
> `MDB_NEXT` operation (i.e. we've reached the end of DB), but keys which
> popped up previously are perfectly valid, and we're just ignoring them here?
> (shouldn't it read `rc != 0 && rc != MDB_NOTFOUND`, just like above?)
Yes, this is indeed wrong - it only works if the loop is left via 'if
(!arr.startsWith(prefix)) break`, but not when all entries in the key range
have a matching prefix.
If an additional `rc` check here is correct is hard to say. We **are** in an
error condition when `rc != {NOTFOUND, SUCCESS}`, but all the values in
`termIterators` are valid.
I vote for removing the rc check here again.
REPOSITORY
R293 Baloo
REVISION DETAIL
https://phabricator.kde.org/D18664
To: valeriymalov, #baloo, bruns, poboiko
Cc: cullmann, ngraham, bruns, kde-frameworks-devel, #baloo, domson,
ashaposhnikov, michaelh, astippich, spoorun, abrahams
