Cherry-pick Allow role names started with 'gp_' and etc #860
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
2 tasks
|
pls rebase to fix ci . |
yjhjstz
reviewed
Jan 13, 2025
Long log: As the title said, this commit removes dead functions in `src/backend/cdb`. Signed-off-by: Yongtao Huang <[email protected]>
The `width` is now in `joinrel->reltarget` and has been set in `build_joinrel_tlist()` and `add_placeholders_to_joinrel()`. After the merge of PostgreSQL 9.6, the concept of `PlaceHolderVar` was introduced, calculating in this way may cause us to ignore the width of `PlaceHolderVar` and thus underestimate the cost of joinrel.
PlanRowMark.canOptSelectLockingClause field is removed as it has been replaced by RangeTblEntry.rellockmode field and never used again.
Authored-by: Bradford D. Boyle <[email protected]>
The comment seems to have been incorrectly handled while dealing with backports and the Postgres 9.4 merge. Update it to what it is in Postgres REL_12_STABLE as that should be the correct comment.
Long log: Don't need to cast the unknown typed literal to text. Drop the cast and remove FIXME. Related commit: 1e7c4bb Signed-off-by: Yongtao Huang <[email protected]>
As part of https://github.com/greenplum-db/gpdb/pull/15187, we added partial wal file tests to gpdb6. Since partial wal file is already supported by gpdb7, we just need to port the tap tests from the aforementioned PR. Partial wal scenario: A few previous commits de76884 and 7cbee7c introduced the concept of creating and archiving partial wal files when the mirror is promoted while the primary is still alive. In this case, the last wal file on the old timeline in the mirror's pg_wal dir is renamed with the suffix ".partial". This partial file also get archived. The original wal file only gets archived once the user runs pg_rewind. Consider the following example: Let's assume that 0000100004 is the current wal file on the master start with a master and standby pair add data to master contents of pg_wal on master: 0000100001 ..... 0000100003 0000100004 - current wal file on master master is alive but standby gets promoted contents of pg_wal on standby: 0000100001 .... 0000100003 0000100004.partial (note that 0000100004 does not exist on the standby) 0000200004 - current wal file on standby Contents of the archive location: 0000100003.tar.gz 0000100004.partial.tar.gz stop master with pg_ctl stop -m fast contents of pg_wal on master: 0000100004 on the master gets flushed and gets archived 0000100004.done gets created on master Contents of the archive location: 0000100003.tar.gz 0000100004.partial.tar.gz 0000100004.tar.gz run pg_rewind copies from standby to master removes 0000100004 and 0000100004.done from master's pg_wal dir Original commit: 5a2975e20bc Original Author: kaknikhil
A previous commit 700cad0 added tap tests for the partial wal file feature. This test was a bit flaky on the ci pipeline. Context: Once we promote the standby(while the primary is running), a partial wal file gets generated on the standby. In order to test if this file gets generated with the correct name, we were previously querying the pg_stat_archiver view. We would compare the name of the last archived wal with the expected partial wal filename. It's possible that this comparison would return wrong results on some platforms because of the ">=" predicate(which could change based on the sorting order.) Fix: This commit removes the use of pg_stat_archiver view and instead relies on the pg_stat_file function to check for the existence of the partial wal file and other files in the tap test
Commit 700cad0 add 'has_streaming => 1' when make a standby node. This will start a walreceiver process to receive the wal log from primary node, rather than just store its contents on the archives of the primary. In this way, the inserted 2000 rows can all have been present on standby node. Revert the enable streaming change when create standby, append conn info to the postgresql.conf after pass the first check and restart standby node.
Commit bc08de1cc5bb introduced ONLY keyword for GRANT/REVOKE. However there's an issue in the grammar where if the table list starts with a table with ONLY, it could not contain more tables with ONLY. Fixing it and adding test cases for syntax coverage.
The merge with upstream absorbed 2930078 which disallowed creating roles with reserved names because some default system roles started with 'pg_' were introduced since 9.6 so we don't want them to conflict: https://www.postgresql.org/docs/9.6/default-roles.html However, in Greenplum reserved names include not only 'pg_' but also 'gp_'. But we do not have any system roles starting with 'gp_' so it is fine to allow users to still create those roles. Now modifying IsReservedName() to not check 'gp_' but adding a new function isReservedGpName() to do that. This avoids the need to modify the signature of IsReservedName() also helps avoid the same issue in future where 'gp_' is banned for existing type of object names which affect existing users. Fix #15259 # Conflicts: # src/backend/catalog/catalog.c
This FIXME was for planner's lack of partition pruning in certain cases. The logic in GP6 could generate wrong results, so it was decided to remove the over-eager optimization for GP7. Orca never did pruning in this case. See https://github.com/greenplum-db/gpdb/issues/10287 and https://github.com/greenplum-db/gpdb/pull/14553 for additional context.
Attnum-filenum mapping was introduced in pg_attribute_encoding in the commit cdd03c165ad The mapping represents the range of files used by the column for an AOCO table. This change is to fix pg_aocsseg table to work show the correct physical segno using this filenum field Co-authored-by: Soumyadeep Chakraborty <[email protected]>
The new attnum-filenum changes on pg_attribute_encoding introduced in cdd03c165ad and the column rewrite operations introduced in last commit changes the range of aocs files on disk. The column filename is now dependent on the filenums and not attnums. This commit changes how we look for possible aocs column files on disk for a table Earlier we used to only have files in order based on natts. Now, as the column can be rewritten we also need to check the filenum pair (i, i+MaxHeapAttributeNumber) of every filenum. Co-authored-by: Soumyadeep Chakraborty <[email protected]> Co-authored-by: Ashwin Agrawal <[email protected]> Co-authored-by: Huansong Fu <[email protected]>
Tao-Ma
force-pushed
the
merge/gpdb_commits
branch
from
993fe9d to
76bea20
Compare
yjhjstz
approved these changes
Jan 14, 2025
my-ship-it
approved these changes
Jan 14, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #ISSUE_Number
What does this PR do?
Type of Change
Breaking Changes
Test Plan
make installcheckmake -C src/test installcheck-cbdb-parallelImpact
Performance:
User-facing changes:
Dependencies:
Checklist
Additional Context
CI Skip Instructions