configure-combinations failed on RHEL 8.0

Host: ec2-52-86-208-247.compute-1.amazonaws.com
Project: WiredTiger (develop)
Commit: diff: WT-6086 Create structures and helpers for time windows. (#5576)

• Create structures and helpers for time windows.

Including both time windows and aggregated time windows.

• Take the next step - there is more to go.

• Get time windows and aggregates compiling.

• Remove extra copy of the aggregate information from the checkpoint structure, fix the checkpoint
  load and salvage code to use the checkpoint's WT_TIME_AGGREGATE structure.

• clang-format

• Set the time window references on function entry, replace local time window dump with the
  standard function.

• Sort validity window fields.

• Switch the prepare field to a uint8_t, bools are 4B on Windows.

• Switch the prepare field to a uint8_t, bools are 4B on Windows.

• Don't use a sentence fragment in time window messages.

• Use structure assignment for time window copies, modern compilers get this right.

Simplify window content functions, don't need if/else.

• Inline __wt_rec_addr_ts_update_window() and __wt_rec_addr_ts_update(), they're now
  single line calls to other functions with reasonable names.

• Avoid doing initialization work on the copy structure until it's actually needed, initializing
  time window structures is going to be expensive.

• Fix NULL indirection, don't initialize the time window unless we have to.

• Revert change to move base timestamp values into timestamp.h, they belong in txn.h.

• Remove a comment, this change is correct: there's a bug fix here, the previous code was using
  the wrong values from the unpacked cell for an address type cell.

• Remove a comment, comparing the full time window signature makes sense to me, we shouldn't have
  been able to RLE encode items if they don't have an exactly matching time window.

• There's a reason we initialize aggregate durability information, both newest_start_durable_ts
  and newest_stop_durable_ts to WT_TS_NONE, but I don't remember why. Add placeholder comments,
  it will get fixed in review.

• The durable timestamp names in the metadata files do not match the names in the WT_TIME_AGGREGATE
  structure. That needs comments or updates/backports, I'm not sure which. Flag in the code and it
  will get fixed in review.

• Fix a spelling complaint.

• Update.

• Overflow keys don't need time windows, only values have time information.

• Overflow keys don't need time windows, only values have time information.

• Minor comment wordsmithing, and explain why prepare is included in the time windows.

• Add a check for prepare in time window is empty function.

• Add a comment to explain the reason behind aggregated durable timestamp

• Use same durable timestamp names across metafile also

• Remove a TODO, the work has been committed.

• Address review comments.

• Rename WT_CELL_UNPACK_TIME_PAIRS_CLEARED to WT_CELL_UNPACK_TIME_WINDOW_CLEARED.

• History store is the only remaining user of WT_TIME_PAIR, make it local and rename for clarity.

• __wt_time_aggregate_to_string() wasn't displaying the newest start durable timestamp: add it,
  and re-order the structures and all timestamp service routines to use the same order of time
  window fields.

• The timestamp string array declarations were incorrectly specified, fix them.
  Use the timestamp string array elements in order for clarity.

• Don't displayed a prepared string unless it's actually prepared, it's not a common condition
  and it takes up too much page real estate.

Co-authored-by: Keith Bostic <keith.bostic@mongodb.com>
Co-authored-by: Haribabu Kommi <haribabu.kommi@mongodb.com> | 05 May 20 21:33 UTC
Evergreen Subscription: ; Evergreen Event:

Task Logs