Skip to content

Sync 14x b13 test#2611

Draft
RekGRpth wants to merge 946 commits into
adb-8.xfrom
sync-14x-b13-test
Draft

Sync 14x b13 test#2611
RekGRpth wants to merge 946 commits into
adb-8.xfrom
sync-14x-b13-test

Conversation

@RekGRpth

Copy link
Copy Markdown
Member

No description provided.

RekGRpth and others added 30 commits May 29, 2026 14:10
Commit dee663f added a new SyncRequestHandler sync_handler argument, to
the SimpleLruInit function definition in src/include/access/slru.h. Add
this in GPDB-specific places.
Commit 5028981 added a new function, ATExecCookedColumnDefault, to
src/backend/commands/tablecmds.c, which calls the StoreAttrDefault
function. However, the earlier commit 19cd1cf had already added the new
arguments bool *cookedMissingVal, Datum *missingval_p, and bool
*missingIsNull_p.
…2626)

Commit 8e1f37c changed the generation rule for fmgroids.h macros.
Change in GPDB-specific places.
….cpp (#2627)

Commit 1375422 in src/include/nodes/plannodes.h removed the
resultRelIndex field from the PlannedStmt structure. Remove this in
GPDB-specific code.
Commit be9788e in src/backend/replication/syncrep.c in the
SyncRepWaitForLSN function moved the conditional return higher, adding
additional conditions and a large comment. Commit e174f69 added an
InterruptHoldoffCount assertion before it. However, the earlier commit
6b0e52b had already added a debug log in the same location, and commit
33a6433 added an assertion before it, and commit bdb1ea1 removed
InterruptHoldoffCount assertion.
1) Commit 808e13b in src/backend/storage/file/buffile.c added a new
conditional error message to the BufFileSeek function. However, earlier
commit 5028054 had already done the same, but with incorrect
indentation, and also added a conditional call to the BufFileFlush
function.

2) Commit 808e13b added a new function BufFileTruncateShared to
src/backend/storage/file/buffile.c, while earlier GPDB-specific commits
had already added the functions BufFileGetFilename, BufFileSuspend,
BufFileResume, BufFilePledgeSequential, BufFileStartCompression,
BufFileDumpCompressedBuffer, BufFileEndCompression,
BufFileLoadCompressedBuffer in the same location.
Commit 5028981 in src/backend/parser/parse_utilcmd.c removed the
inh_indexes field from the CreateStmtContext structure, although it was
used in the transformDistributedBy function in the GPDB. Restore it for
the GPDB and add code to populate it in the transformTableLikeClause
function from 7X.
1) Commit 16fa9b2 in the file src/include/catalog/pg_proc.dat added a
new OID 3435, while the earlier commit 19cd1cf already used this OID in
the file src/include/catalog/pg_am.dat.

2) Commit 36b9312 prohibited the use of custom OID symbols in
pg_proc.dat.
Commit 9796f45 in src/bin/pgbench/pgbench.c in the initCreateTables
function removed the local variable opts, replacing snprintf opts with
appendPQExpBuffer query. Replace in GPDB-specific code.
Commit 06a7c31 has added a new is-bare-label
argument to PG_KEYWORD. Most keywords have BARE_LABEL, allowing them to be used
as labels without AS keyword, but 39 are marked AS_LABEL, requiring as. However,
there are a lot of GPDB-specific keywords too, most of them without comments.
Conservatively mark them AS_LABEL to avoid accidentally breaking the grammar,
Also mark all GPDB-specific keywords with comments to simplify future conflict
resolutions. However, some of the keywords, specifically "format" and "ignore",
are present in later versions of Postgres and "format" is marked as BARE_LABEL,
so mark it BARE_LABEL here too.
Commit 19f5a37 simplified code by replacing
type casts with cstmt, however the code nearby was modified by GPDB-specific
commits like c7649f1.
Commit 9796f45 in src/bin/pgbench/pgbench.c in the initCreateTables
function removed the local variable opts, replacing snprintf opts with
appendPQExpBuffer query. Replace in GPDB-specific code.
silent-observer and others added 30 commits June 11, 2026 12:44
Commit 3d351d9 redefined pg_class.reltuples
to be -1 before the first VACUUM or ANALYZE. Adapt the output of
GPDB-specific tests previously missed.
#2674)

Commit cd6f479 added the refobjversion column to the pg_depend catalog
table in src/include/catalog/pg_depend.h. Add it to the GPDB-specific
tests.
…#2673)

1) Commit 9fc2122 in src/backend/commands/indexcmds.c changed the error
message in the DefineIndex function. Change this in the GPDB-specific
test outputs.

2) Commit 6b2c4e5 in src/backend/partitioning/partbounds.c changed the
error position display in the check_new_partition_bound function. Add
the error position display in the GPDB-specific test outputs.
Commit e03e1e1 incorrectly resolved
conflicts related to dumping indexes, partially reverting changes in
getIndexes() function from commit 257836a
because of refactoring in commit 30ab901.
Adjust these changes to the new code structure.
Commit dee663f refactored clog fsyncing logic.
Commit c97c117 uses SYNC_HANDLER_CLOG for
distributed clog syncing, which is not valid since distributed clog have their
own directory.
Fix it by adding a separate handler for distributed clog.
#2692)

Commit c5f42da fixed documentation for
HAVING clause, however GPDB has different WINDOW clause for some reason:
it is missing completely in select_into.sgml and it is missing "[, ...]"
in select.sgml. The actual gram.y indicates WINDOW clause is present both
in SELECT and SELECT INTO, and that WINDOW specifications can be repeated,
so resolve in favor of upstream.
Commit f0f13a3 changed the output rows
estimate for UPDATE without RETURNING to 0. Most tests weren't affected,
since single-line EXPLAIN ignores cost diffs anyway. This test seems to be
the only multi-line EXPLAIN in the test suite where these diffs matter.
Commit be9788e added a fast path for
SyncRepWaitForLSN, but it relies on the synchronous_standby_names GUC, which is
always empty on the coordinator.
Fix it by disabling the fast path on the coordinator.
Commit be9788e added a fast path for
SyncRepWaitForLSN, but it relies on the synchronous_standby_names GUC, which is
always empty on the coordinator.
Fix it by disabling the fast path on the coordinator.
Commit 7b28913 added new TAP test, however
GPDB does not support PREPARE TRANSACTION, so disable that part of the test.
…rimary keys" scenario

This test uses a postfix operator that is no longer supported.
 Fix it by using another operator type.
This error is caused by the execution logic of Hash Join Node, when the outer
node does not fetch a tuple, it calls the ExecSquelchNode method let the lower
nodes not to produce tuples anymore, leaving the inner node "squalched" but
"un-executed" state. If the hash join needs to be rescanned, then the inner nodes
has to be rescanned too if Hash Node's lefttree does not have chgParam, and finally
a node does not have chgParam but "squelched" will cause ERROR.

This commit resolves the issue by handling of the rescan method in HashJoin Node properly.

Authored-by: wuyuhao28 <wuyuhao28@github.com>
(cherry picked from commit 73517b3)
This error is caused by the execution logic of Hash Join Node, when the outer
node does not fetch a tuple, it calls the ExecSquelchNode method let the lower
nodes not to produce tuples anymore, leaving the inner node "squalched" but
"un-executed" state. If the hash join needs to be rescanned, then the inner nodes
has to be rescanned too if Hash Node's lefttree does not have chgParam, and finally
a node does not have chgParam but "squelched" will cause ERROR.

This commit resolves the issue by handling of the rescan method in HashJoin Node properly.

Authored-by: wuyuhao28 <wuyuhao28@github.com>
(cherry picked from commit 73517b3)
…rimary keys" scenario (#2701)

This test uses a postfix operator that is no longer supported.
Fix it by using another operator type.
Commit 257836a added restoring of the collation
versions, but the ggdb modification on the catalog tables requires enabling
allow_system_table_mods GUC.
Commit e03e1e1 incorrectly resolved
conflicts related to dumping indexes, partially reverting changes in
getIndexes() function from commit 257836a
because of refactoring in commit 30ab901.
Adjust these changes to the new code structure.
Commit 5028981 in src/backend/parser/parse_utilcmd.c removed the
inh_indexes field from the CreateStmtContext structure, although it was
used in the transformDistributedBy function in the GPDB. Restore it for
the GPDB and add code to populate it in the transformTableLikeClause
function from 7X.
Commit 5d1833f statrted to use be_tls_* API for SSL information in sslinfo
in extensions functions. It trims expected string produced by
X509_get_subject_name at NAMEDATALEN (64) length. Vanila postgres does not have
tests for sslinfo at all, they were introduced in GPDB. Since the behaviour is documented,
we adjust GPDB tests.
Commit ce90f07 changed error message.
Change ggdb specific tests accordingly.
Commit 257836a added restoring of the collation
versions, but in the ggdb modification of the catalog tables requires enabling
allow_system_table_mods GUC.
Commit 76f412a removed factorial operators,
leaving only the factorial() function. Replace in GPDB-specific tests.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

7 participants