summaryrefslogtreecommitdiff
path: root/src/backend/jit
diff options
context:
space:
mode:
authorAlvaro Herrera <alvherre@alvh.no-ip.org>2022-10-19 10:35:53 +0200
committerAlvaro Herrera <alvherre@alvh.no-ip.org>2022-10-19 10:37:06 +0200
commitdf3737a651f40a398c14348b0482e97031e7a7d4 (patch)
tree69ee6838c88bd709fd00de1edb5263db48079399 /src/backend/jit
parent9668c4a6618cc4a7a94f4bc0ad5685dd596bab63 (diff)
downloadpostgresql-df3737a651f40a398c14348b0482e97031e7a7d4.tar.gz
Remove pg_backup_start_callback and reuse similar code
We had two copies of almost identical logic to revert shared memory state when a running backup aborts; we can remove pg_backup_start_callback if we adapt do_pg_abort_backup so that it can be used for this purpose too. However, in order for this to work, we have to repurpose the flag passed to do_pg_abort_backup. It used to indicate whether to throw a warning (and the only caller always passed true). It now indicates whether the callback is being called at start time (in which case the session backup state is known not to have been set to RUNNING yet, so action is always taken) or shmem time (in which case action is only taken if the session backup state is RUNNING). Thus the meaning of the flag is no longer superfluous, but it's actually quite critical to get right. I (Álvaro) chose to change the polarity and the code flow re. the flag from what Bharath submitted, for coding clarity. Co-authored-by: Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com> Discussion: https://www.postgresql.org/message-id/20221013111330.564fk5tkwe3ha77l%40alvherre.pgsql
Diffstat (limited to 'src/backend/jit')
0 files changed, 0 insertions, 0 deletions