GitLab

  in this version. Sure enough, it never caused any improvement in
  the execution speed and rather caused a small increase of execution
  time. This is probably because values are sorted by rowid in each
  range of CONNECT indexes. This could be reconsidered if a customer
  have a need for processing very big files.

- Fix a bug in ha_connect::CheckCond. The negated form of BETWEEN and
  IS NULL operators was not recognized.
modified:
  storage/connect/ha_connect.cc

- Add long jump initialization in CntReadNext. This was causing a server
  crash when an error occured in a ReadColumn.
modified:
  storage/connect/connect.cc

- General cleanup of CONNECT source code eliminating all code not used by
  CONNECT, including the MRR test code (saved separately).
modified:
  storage/connect/catalog.h
  storage/connect/colblk.cpp
  storage/connect/colblk.h
  storage/connect/connect.cc
  storage/connect/connect.h
  storage/connect/domdoc.h
  storage...