-
Type: Task
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: None
In WT-11297, I slightly modified an existing piece of code in cur_backup.c:curbackup_close for incremental backup to checkpoint the metadata file specifically at the end of backup to get the correct, new ID strings into the metadata.
The code was unchanged from code we execute on when incremental=(force_stop=true) is used other than it was used on a normal incremental backup cursor close. Running timestamp_abort with backup on, that draft PR change resulted in data mismatches on the source database.
We should add testing for force_stop=true to timestamp_abort to test that code path more fully.