> On Oct. 14, 2014, 1:11 a.m., Parth Chandra wrote: > > contrib/native/client/src/clientlib/drillClientImpl.cpp, line 802 > > <https://reviews.apache.org/r/26672/diff/1/?file=719859#file719859line802> > > > > Did you just leak the column defs?
It gets cleaned up in clearAndDestroy. The issue was that because after each record batch, prevSchema got cleaned up and since prevSchema is the same as m_columnDefs, m_columnDefs was always empty when the next record batch came in. Because of this, every subsequent record batch would set the hasSchemaChanged flag to true as long as it wasn't empty. - Norris ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/26672/#review56478 ----------------------------------------------------------- On Oct. 14, 2014, 12:41 a.m., Norris Lee wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/26672/ > ----------------------------------------------------------- > > (Updated Oct. 14, 2014, 12:41 a.m.) > > > Review request for drill and Parth Chandra. > > > Repository: drill-git > > > Description > ------- > > hasSchemaChanged is always set as true for all record batches except the > first one regardless of whether the schema has changed or not, including > cases where specific columns are projected, which should never happen. > > > Diffs > ----- > > contrib/native/client/src/clientlib/drillClientImpl.cpp 77795ed > > Diff: https://reviews.apache.org/r/26672/diff/ > > > Testing > ------- > > > Thanks, > > Norris Lee > >
