Home - Waterfall Grid T-Grid Console Builders Recent Builds Buildslaves Changesources - JSON API - About

Console View


Categories: connectors experimental galera main
Legend:   Passed Failed Warnings Failed Again Running Exception Offline No data

connectors experimental galera main
fe03bd6e9acf...
fe03bd6e9acf5f86fccb8bf047764535e418d5bb
Daniel Black
deb: add Ubuntu stonking as next release
fc0ee7030d5b...
fc0ee7030d5b07c38f77841f1779006a060af0ac
Dave Gosselin
MDEV-28509:  Dereferenced null pointer of type 'struct JOIN_TAB' in add_key_field

setup_group no longer writes Item::marker.  The ONLY_FULL_GROUP_BY
check now tests GROUP BY membership by walking the GROUP BY list.

A query that defines a WINDOW but never refers to it could crash in
add_key_field, for example

  WITH cte AS (SELECT i FROM (SELECT i FROM t1 GROUP BY i) dt
              WINDOW w AS (PARTITION BY i))
  SELECT a.i FROM cte a JOIN cte b ON a.i=b.i WHERE a.i != 5;

A query that defines a WINDOW goes through setup_group, which set
marker to MARKER_UNDEF_POS (-1) on each GROUP BY expression so that
the ONLY_FULL_GROUP_BY check could skip it.  Other code reads marker
as a set of flag bits (-1 sets all bits).
Item_direct_view_ref::grouping_field_transformer_for_where then took
the ref as flagged for substitution and followed a path that ends in
the crash.

The ONLY_FULL_GROUP_BY check was the only reader of that value, so
MARKER_UNDEF_POS is removed.  The necessary check is local to the
setup_group function.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
f8d9e3481c97...
f8d9e3481c970762286fb8f646c197cd82494d5a
Marko Mäkelä
fixup! 02a56c4d7d22678ba826c5c3a5e4849b05a7ae4f
ee1326a42d66...
ee1326a42d667ea91a37c94152f6fc4f43d68069
Oleksandr Byelkin
MDEV-41186 Fix stack-buffer-overflow in make_unique_constraint_name

A multi-byte PERIOD name at the 64-character limit (192 bytes) filled
the name buffer exactly, leaving no room for the '_N' suffix appended
when generating a unique name for the implicit CHECK constraint.

Truncate on a character boundary, but only once a suffix is actually
needed (mirroring make_unique_key_name), so a non-colliding name is
never altered.

Co-Authored-By: Claude Sonnet 5 <[email protected]>
ed38ffc01b84...
ed38ffc01b849c7a491d8ca7e4f606b4492c6dc9
Vladislav Vaintroub
MDEV-40967 PROXY protocol host check sent in clear text mid-SSL handshake

Defer the host-privileged/host-blocked check for a PROXY-header-derived
address until after the client's SSL handshake completes, instead of
sending it immediately in clear text.

Co-Authored-By: Claude Sonnet 5 <[email protected]>
e036579a5a69...
e036579a5a69f6d388bfa26f8a6967b7554d3711
Aleksey Midenkov
MDEV-41157 CREATE DATABASE COMMENT overflows db.opt comment buffer

Bug 1: put_dbopt() used strmov() to copy schema_comment into a fixed
DATABASE_COMMENT_MAXLEN+1 buffer. validate_comment_length() only
truncates comment->length in non-strict sql_mode, leaving comment->str
NUL-terminated at its original (unbounded) length. strmov() copies
until the source NUL, ignoring the truncated length, overflowing the
destination buffer for long comments.

The fix uses strmake() bounded by comment->length instead, matching
the LEX_CSTRING contract (length is authoritative, str need not be
NUL-terminated at length).

Bug 2: write_db_opt() used strxnmov() to copy the full un-truncated
comment until it ran out of buffer space mid-string with no trailing
newline. Which made load_db_opt() silently discard the whole
unterminated "comment=" line on the next restart, losing the comment
entirely instead of just truncating it.

The fix bounds the comment copy into db.opt by the already-validated
comment->length via strmake(), instead of relying on the source
string's own NUL terminator, matching the put_dbopt() fix.

Bug 3: validate_comment_length() only runs on a COMMENT clause given
in the current statement. ALTER DATABASE without one instead pulls
the existing comment off disk via load_db_opt(), which never bounded
it. That unvalidated length then reached write_db_opt()'s own
comment= copy into its stack buffer, so a legacy or hand-edited
db.opt with an overlong comment= line overflowed it on ALTER DATABASE.

The fix: load_db_opt() now clamps the parsed comment to
DATABASE_COMMENT_MAXLEN right when it reads the "comment=" line, so
every consumer (put_dbopt(), write_db_opt()'s ALTER path) always sees
an already-bounded value.

The clamp itself must truncate by bytes, not characters:
Well_formed_prefix()'s LEX_CSTRING overload takes a character count,
but DATABASE_COMMENT_MAXLEN sizes the buffers in bytes.

Bug 4: validate_comment_length() had the same character/byte
confusion, on the primary CREATE/ALTER DATABASE COMMENT path: it
passed max_len as a character count to Well_formed_prefix(), so a
multi-byte comment could be truncated to max_len characters instead of
bytes.

Fixed like load_db_opt(): clamp to max_len bytes first, then find the
well-formed prefix.
dd1c1e3d14a2...
dd1c1e3d14a28009c33633f35e725d03aa31d981
Sergei Golubchik
MDEV-40229 I_S.VECTOR_INDEXES: Unclear or inconsistent semantics of INDEX_SIZE

11.8 fix for I_S.TABLES.INDEX_LENGTH column

Take into account both data_file_length and index_file_length,
they both take space and both belong to the vercor index.
d49a7da399c7...
d49a7da399c707842f6e2690f5afbb378c4b1585
Sergei Petrunia
Add comment about Create_tmp_table::m_group
d30527240832...
d305272408324fd676e2b7dd56cce6e936d11773
Thirunarayanan Balathandayuthapani
MDEV-19574: innodb_stats_method is not honored when innodb_stats_persistent=ON

Problem:
=======
When persistent statistics are enabled (innodb_stats_persistent=ON),
the innodb_stats_method setting is not properly utilized during
statistics calculation.

The statistics collection functions always use a hardcoded default
behavior for NULL value comparison instead of respecting the
configured stats method. This affects the accuracy of
n_diff_key_vals (distinct key count), particularly for
indexes with nullable columns containing NULL values.

Moreover, stat_n_non_null_key_vals[] was never computed for
persistent statistics; it stayed at the 0 that
dict_stats_empty_index() assigns.

With innodb_stats_method=nulls_ignored, innodb_rec_per_key()
therefore always found n_diff <= n_null and reported one record
per key for every index. This impacts the query optimizer,
which makes decisions based on inaccurate cardinality estimates.

Solution:
========
Introduced IndexLevelStats to collect statistics at a specific
B-tree level during index analysis.

Introduced PageStats to collect statistics for leaf page analysis.

Refactored the following functions:
dict_stats_analyze_index_level() to IndexLevelStats::analyze_level()
dict_stats_analyze_index_for_n_prefix() to IndexLevelStats::sample_leaf_pages()
dict_stats_analyze_index_below_cur() to PageStats::scan_below()
dict_stats_scan_page() to PageStats::scan()

The innodb_stats_method value is read once per table in
dict_stats_update_persistent() and passed down, so that all
indexes of a table are analyzed with the same method.

Add the stats method name to stat_description when
innodb_stats_method has a non-default value. The suffix is
dropped when the description is already full.

Added the new stat name n_nonnull_fld01, n_nonnull_fld02, etc.
with a stats description, to indicate how many non-null values
exist for the nth field of the index. This value is retrieved
and stored in the index statistics
in dict_stats_fetch_index_stats_step(). The counts are per
column, not per n-column prefix.

rec_get_n_blob_pages(): Calculate the number of
externally stored pages for a record, using ceiling division
by the usable BLOB page payload (blob_part_size), which differs
between ROW_FORMAT=COMPRESSED (zip_size minus FIL_PAGE_DATA) and
the other formats (srv_page_size minus the BLOB header and the
page trailer). For ROW_FORMAT=COMPRESSED the length in
the field reference is the uncompressed length, so the result is an
upper bound.

When the leaf level is scanned in full, the number of leaf pages that
were scanned is reported as n_leaf_pages for a multi level index.
Before, result.n_leaf_pages was overwritten with
index->stat_n_leaf_pages, which dict_stats_empty_index() had
just set to 1, so every index that took the full scan path
reported n_leaf_pages=1. Single page indexes report 1.
This changes cardinality estimates and
therefore leads to multiple changes in existing test cases.

Non-null values are counted only at the leaf level, since only leaf
pages hold actual records. A full scan of the leaf level counts them
exactly. When the level is sampled, the per column count is derived
from the sampled leaves with the same formula as n_diff:

  n_ordinary_leaf_pages * n_non_null_all_analyzed_pages
                        / n_leaf_pages_to_analyze

This is an estimate for NOT NULL columns as well: the sampled leaves
may hold fewer or more records than the average, and a dive that
stops at a boring page contributes nothing to the sum while still
counting in the divisor.

innodb_rec_per_key(): stat_n_non_null_key_vals[i] holds the
number of records in which the i-th indexed column alone is
not NULL, while what has to be excluded here is the number
of records whose first i+1 columns are all not NULL,
because that is the population which the n-column
prefix statistic stat_n_diff_key_vals[i] has to be
corrected against when innodb_stats_method=nulls_ignored:
with NULLs compared as unequal, every record carrying a NULL
anywhere in the prefix adds a distinct value of its own to n_diff.

PageStats::scan(): n_non_null is accumulated and assigned only
for leaf pages, so that a non-leaf scan cannot leave a node
pointer count behind when scan_below() stops at a boring page
without reaching a leaf.

IndexLevelStats::reset_for_level() also clears n_diff[], and
dict_stats_analyze_index() zero initializes the buffer backing it, so
that a level scan which finds no records (a failed
btr_pcur_open_level(), or a non-leaf page whose first record is not
marked as the leftmost one on the level) leaves n_diff[] at 0 instead
of stale values.

IndexLevelStats::sample_leaf_pages() returns early when the group
boundaries for the prefix are empty, which is the same condition.

IndexLevelStats::analyze_level(): Instead of copying the last record
of the page, retain the latch on the page until the record has been
compared with the first record of the next page

dict_stats_fetch_index_stats_step() no longer resets
stat_n_non_null_key_vals[] while processing an n_diff_pfxNN row:
dict_stats_empty_table() has already cleared the array before the
fetch, and with n_nonnull_fldNN rows now being read too,
that reset would make the result depend on the order in which the
rows arrive.

dict_stats_save(): now static function in dict0stats.cc that
takes the innodb_stats_method value, and is removed from dict0stats.h.
dict_stats_update_persistent() saves the statistics itself, so its
callers no longer have to.

Replaced btr_rec_get_externally_stored_len() with
rec_get_n_blob_pages() in dict0stats.cc.

btr_rec_get_field_ref_offs() and btr_rec_get_field_ref(),
together with the BTR_BLOB_HDR_* macros, were moved from btr0cur.cc
to btr0cur.h so that rec_get_n_blob_pages() can reuse them;

btr_rec_get_field_ref_offs() is now a noexcept function
returning size_t.

Changed stat_n_diff_key_vals and stat_n_non_null_key_vals from
ib_uint64_t* to uint64_t*

len_is_stored(): simplified to a single comparison, which is
equivalent for the unsigned lengths that it is used with.

Removed the unused UNIV_STATS_DEBUG build macro (univ.i) and
turned the DEBUG_PRINTF() helper in dict0stats.cc into an
unconditional no-op
cf2d867ef08b...
cf2d867ef08b5b5326060a13f62d4088401910d5
Vladislav Vaintroub
fix macOS: -undefined dynamic_lookup instead of linking mariadbd

mariadbd is not exported, so it can't be a dependency of mariadb_private
(previous commit dropped that link entirely). On Apple platforms
specifically, -bundle_loader isn't the only option - a MODULE there is
a loadable bundle, same class as Python/Perl/Ruby native extensions,
and -Wl,-undefined,dynamic_lookup is the standard way those defer
symbol resolution to load time, same as ELF already does for free.
CMake's own Platform/Darwin.cmake confirms this isn't the default for
MODULE targets, so it needs to be added explicitly.

Scoped to APPLE specifically, not "not Linux" - FreeBSD/OpenBSD/NetBSD
and Solaris/illumos are ELF like Linux and never needed anything here.

Assisted-by: Claude:claude-5-sonnet
cea4b3118792...
cea4b3118792148c1051fb0eb5afed4e05b9512f
Khaled Riyad
MDEV-37563 Crash on a FOR loop with a row comparison in the upper bound

The rule for_loop_bound_expr left the items of a FOR loop bound on
THD::free_list, so they were cleaned up after the wrong instruction, and
a row comparison in the upper bound kept a pointer to memory freed at
the end of an instruction.

Store the items in the sp_assignment_lex of the bound, like
assignment_source_expr does, and pass them back to THD::free_list right
before generating the instruction that evaluates that bound.
c5539ea9d9c3...
c5539ea9d9c390ac8fdfd4d2322f2f0713eab29f
Sergei Golubchik
MDEV-40608 MariaDB-devel is incomplete for plugins

This works on Linux and on Windows, with rpm/deb/tar.gz/zip
installations.

For rpm/deb it just works, for tar.gz/zip there is no
standard location, so one needs to configure plugin with

  -DCMAKE_PREFIX_PATH=/pah/to/mariadb/basedir

after that, `cmake --install .` works too, installing in the same
basedir.

`cmake --build . --target package` works, creating rpm/deb/targz/zip
depending on whether it's Linux or Windows and whether -DRPM or -DDEB
was specified.

* create and install mariadb-plugin-config.cmake
* for now it only supports one plugin per project, error out
  if there are many
* deb: move all headers that plugins need to libmariadb-dev,
  together with libmysqlservices.a. At least until we'll
  create mariadb-plugin-dev. Nobody should need huge
  libmariadbd-dev to develop a plugin
* rpm: all in MariaDB-devel already, no changes here
* install wsrep headers too, THD layout depends on WITH_WSREP
* show DBUG_OFF, ENABLED_DEBUG_SYNC, and SAFE_MUTEX to plugins, same
  reason (it doesn't happen automatically as they're not in my_config.h)
* but don't install config.h - high chance of name conflict with other
  projects and it's an exact copy of my_config.h anyway.
* adjust plugin.cmake to work for external plugins
* move server-internal part to top-level CMakeLists.txt
* remove double-defined macros from unireg.h (the guard doesn't help
  if unireg.h is included first)
* package  plugin metadata as yaml in .tar.gz/.zip

ColumnStore, until fixed, needs a backward-compatibility workaround
c2314ca1c7a0...
c2314ca1c7a09bdb7b05a0147c3e98cef0c23489
Raghunandan Bhat
MDEV-41193: ASAN heap-buffer-overflow in ha_connect::CheckCond after select from Connect table

Problem:
  When CONNECT engine pushes a WHERE clause down to an external table,
  it writes the filter into the work area, without checking how much
  space is left. A large string literal in the WHERE clause can overflow
  the work area allocated by the engine. For ex: if connect_work_size is
  set to 4MB and the string literal in the WHERE clause is larger than
  4MB, it can grow past the allocated work area.

Fix:
  Track the space left in the work area and check it before writing. If
  the filter doesn't fit, drop it instead of writing past the buffer.
be67a4fc1e04...
be67a4fc1e0493913732df90e562f122bff9dfe3
Georg Richter
CONC-854: Fix regression when disable-ssl-verify-server-cert is used with ssl_ca

The changes in commit e8c0a16d caused disable-ssl-verify-server-cert
(tls_allow_invalid_server_cert) to be ignored whenever ssl_ca,
ssl_capath, ssl_crl, or ssl_crlpath were set.

While intended as a security measure, this introduced a breaking change
for IP-addressed connections (such as 127.0.0.1 and ::1) and environments
where clients rely on ssl_ca for CA chain validation while explicitly
disabling hostname/IP verification (including MTR test suites).

This patch partially reverts e8c0a16d in ma_tls.c:
- Restores explicit precedence to tls_allow_invalid_server_cert so
  disabling server verification bypasses peer verification regardless of
  whether CA/CRL options are present.
- Clearing mysql->net.tls_verify_status when verification is explicitly
  disabled ensures that cleartext auth-switch guards in my_auth.c continue
  to function correctly for authorized unverified connections.
- Updates unit tests in tls.c.in to reflect the restored behavior for
  MARIADB_TLS_DISABLE_PEER_VERIFICATION when ssl_ca is configured.
  • cc-x-codbc-windows: 'dojob pwd if '3.4' == '3.4' ls win32/test SET TEST_DSN=master SET TEST_DRIVER=master SET TEST_PORT=3306 SET TEST_SCHEMA=odbcmaster if '3.4' == '3.4' cd win32/test if '3.4' == '3.4' ctest --output-on-failure' failed -  stdio
bd64156fd877...
bd64156fd877c13502dd8aeedd8f7ffc1cbd53a9
Vladislav Vaintroub
fix Windows build: split mariadb_private into headers-only and full parts

Split mariadb_private into a headers-only part, linked to static-only
plugins, and one that additionally links the server library (to fix
unresolved dependencies) for module plugins. Fixes the cycle reported
by CMake. Also, rocksdb_aux_lib needs server headers too.

Also fixes INSTALL_RUNTIME_DEPS: it walked only one level of
LINK_LIBRARIES to find same-build shared libs to exclude, which worked
while plugins linked server/mariadbd directly. Now that they link the
mariadb_private INTERFACE library instead, server only shows up via
its INTERFACE_LINK_LIBRARIES, so the walk is now transitive.

Assisted-by: Claude:claude-5-sonnet
b325aea7a726...
b325aea7a7268c0e34b36b855b8ab76decb518c5
Vladislav Vaintroub
MDEV-40967 PROXY protocol host check sent in clear text mid-SSL handshake

Defer the host-privileged/host-blocked check for a PROXY-header-derived
address until after the client's SSL handshake completes, instead of
sending it immediately in clear text.

Co-Authored-By: Claude Sonnet 5 <[email protected]>
af40ebd79f7e...
af40ebd79f7e335d67faabcbd72d8bac4d25eae3
Raghunandan Bhat
MDEV-41193: ASAN heap-buffer-overflow in ha_connect::CheckCond after select from Connect table

Problem:
  When CONNECT engine pushes a WHERE clause down to an external table,
  it writes the filter into the work area, without checking how much
  space is left. A large string literal in the WHERE clause can overflow
  the work area allocated by the engine. For ex: if connect_work_size is
  set to 4MB and the string literal in the WHERE clause is larger than
  4MB, it can grow past the allocated work area.

Fix:
  Track the space left in the work area and check it before writing. If
  the filter doesn't fit, drop it instead of writing past the buffer.
ad39ca01d63c...
ad39ca01d63c33fb0286d749d4a41d468acf8ee1
Vladislav Vaintroub
experiment: drop mariadbd link from mariadb_private on non-MSVC/AIX/Linux

This was causing "install(EXPORT ...) includes target mariadb_private
which requires target mariadbd that is not in any export set" on
macOS, since nothing exports mariadbd. Removing it here to see whether
anything actually needs it in practice - CMake's default MODULE
creation flags on Apple platforms don't add -undefined dynamic_lookup
(checked cmake's own Platform/Darwin.cmake), so if a plugin does
reference symbols outside mysqlservices, this should fail to link
rather than silently misbehave.

Assisted-by: Claude:claude-5-sonnet
abb3891a9e63...
abb3891a9e639128fc4e7ee5a47d592564d5eb29
Brandon Nesterenko
MDEV-40906: rpl.rpl_gtid_thread_id assert_grep.inc failed

rpl.rpl_gtid_thread_id could fail sporadically due to a
non-deterministic slave state during an assert. The test asserted that
a certain number of transaction's exist in the slave's binary log file;
however, there was no sync between the master and slave after the last
transaction executed on the master. This means the slave's binary log
could be checked before the transaction ever was sent to/committed on
the slave.

The fix is to simply sync the master and slave before checking the
slave's binary log.

Signed-off-by: Brandon Nesterenko <[email protected]>
a456c3694a98...
a456c3694a98dad6bfa112d081acb58282a7185a
Aleksey Midenkov
MDEV-41050 versioned DELETE via row_end index leaves a row undeleted

1. Aria/MyISAM engines

A system-versioned DELETE on a MyISAM or Aria table indexed by row_end
could leave the last current row undeleted. The delete scans the current
rows via an equality search on row_end = MAX, which is driven by
mi_rnext_same/maria_rnext_same. That function keeps the search's
reference key in lastkey2 and uses the HA_STATE_RNEXT_SAME flag to
remember it has already stored it. This reference-key mechanism is
exactly what lets the scan modify rows it is walking without skipping
them.

Deleting a versioned row is an in-place update of row_end, and the update
reuses lastkey2 as scratch space for the changed key, so it clears
HA_STATE_RNEXT_SAME to request that rnext_same re-store its reference on
the next call. However, TABLE::delete_row wraps the update in
HA_EXTRA_REMEMBER_POS/HA_EXTRA_RESTORE_POS, and RESTORE_POS restored the
whole saved info->update word, resurrecting the HA_STATE_RNEXT_SAME bit
that the update had just cleared.

As a result rnext_same skipped rebuilding its reference key and compared
subsequent keys against the now-overwritten lastkey2, hitting a spurious
end-of-file and terminating the scan one row early, defeating the engine's
own protection against a Halloween-style skip.

Fixed by preserving the current HA_STATE_RNEXT_SAME bit across
RESTORE_POS instead of restoring the stale saved value. See also the HEAP
fix below: same root cause, different per-engine mechanism.

2. HEAP engine

The row loss also reproduces on the MEMORY (HEAP) engine. A system-
versioned DELETE scans the current rows on the row_end index and turns
each delete into an in-place update of row_end, so it modifies the very
index it is walking.

When the changed key is the scanned one (info->lastinx), hp_delete_key()
repositions the cursor but heap_update() leaves info->update untouched, so
HA_STATE_NEXT_FOUND from the preceding heap_rnext() stays set. The next
heap_rnext() then sees current_ptr == 0 with that bit and takes the
"!current_ptr && HA_STATE_NEXT_FOUND" guard as a false end-of-file,
stopping one row early.

Fixed by clearing HA_STATE_NEXT_FOUND when the scanned index key changed.
HA_STATE_AKTIV is kept (unlike heap_delete): the row is updated, not
removed, so a following op must not fail test_active(). The bit is only
set after a heap_rnext(), so a plain single-row UPDATE never reaches this.
See also the Aria/MyISAM fix above: same root cause, different per-engine
mechanism.

3. Why the fix differs per engine, and InnoDB

Aria and HEAP both trace their handler design back to MyISAM, hence the
same root cause (stale scan bookkeeping after an in-place key change) in
all three, fixed at each engine's own bookkeeping spot. InnoDB needs no
fix: its persistent cursor survives concurrent index modification by
design, already covered by this same test under the timestamp
combination (default-storage-engine=innodb), which passes unmodified.
a44bfd9febdd...
a44bfd9febdd6ebc20073011a0e1bc3cf41d3e54
bsrikanth-mariadb
MDEV-40837: Crash while recording context when character_set_results is NULL

Unlike character_set_client or collation_connection,
character_set_results can be set to NULL. When recording context for
a query in opt_context_store_replay.cc,
character_set_results->cs_name was accessed without first checking
whether character_set_results itself was NULL, causing a crash.

Fix Optimizer_context_recorder::dump_sql_script() to check
character_set_results for NULL before accessing cs_name.

Added a test for the same, in opt_context_store_stats.test
956dd7b04893...
956dd7b048935331c754d5a47edd9eb0f8d55df9
Vladislav Vaintroub
MDEV-40967 PROXY protocol host check sent in clear text mid-SSL handshake

Defer the host-privileged/host-blocked check for a PROXY-header-derived
address until after the client's SSL handshake completes, instead of
sending it immediately in clear text.

Co-Authored-By: Claude Sonnet 5 <[email protected]>
8ef7ed01dc74...
8ef7ed01dc741e674221512032f448b76ffee45c
Vladislav Vaintroub
MDEV-32745 followup - fix Windows build on a machine without Developer Mode.

In custom CMake target, do not attempt symlink with hardlink fallback,
this makes VS build fail since stderr is written (and attempts to suppress
stderr fail in Ninja generator)

Just use hardlinks, no substantial difference in the given scenario.
8565ec28890d...
8565ec28890d31a2f7cfc950e25288b77d296072
Oleksandr Byelkin
MDEV-41175 Fix stack-buffer-overflow in backup_log_ddl()

backup_log_ddl() sized its stack log buffer assuming each
identifier is at most ~40 bytes, but add_name_to_buffer() can
expand each identifier character to 5 bytes when re-encoding it
into my_charset_filename, so a RENAME TABLE with long, special-
character names overflowed the buffer (reported under ASAN).

Fixed by sizing the buffer to the true worst case per identifier,
and by making add_str_to_buffer() and its callers assert on an
explicit end-of-buffer pointer before writing.

Co-Authored-By: Claude Sonnet 5 <[email protected]>
75efad04d5df...
75efad04d5df0089c1e02194ca7aaf5b406372e8
Thirunarayanan Balathandayuthapani
MDEV-19574: innodb_stats_method is not honored when innodb_stats_persistent=ON

Problem:
=======
When persistent statistics are enabled (innodb_stats_persistent=ON),
the innodb_stats_method setting is not properly utilized during
statistics calculation.

The statistics collection functions always use a hardcoded default
behavior for NULL value comparison instead of respecting the
configured stats method. This affects the accuracy of
n_diff_key_vals (distinct key count), particularly for
indexes with nullable columns containing NULL values.

Moreover, stat_n_non_null_key_vals[] was never computed for
persistent statistics; it stayed at the 0 that
dict_stats_empty_index() assigns.

With innodb_stats_method=nulls_ignored, innodb_rec_per_key()
therefore always found n_diff <= n_null and reported one record
per key for every index. This impacts the query optimizer,
which makes decisions based on inaccurate cardinality estimates.

Solution:
========
Introduced IndexLevelStats to collect statistics at a specific
B-tree level during index analysis.

Introduced PageStats to collect statistics for leaf page analysis.

Refactored the following functions:
dict_stats_analyze_index_level() to IndexLevelStats::analyze_level()
dict_stats_analyze_index_for_n_prefix() to IndexLevelStats::sample_leaf_pages()
dict_stats_analyze_index_below_cur() to PageStats::scan_below()
dict_stats_scan_page() to PageStats::scan()

The innodb_stats_method value is read once per table in
dict_stats_update_persistent() and passed down, so that all
indexes of a table are analyzed with the same method.

Add the stats method name to stat_description when
innodb_stats_method has a non-default value. The suffix is
dropped when the description is already full.

Added the new stat name n_nonnull_fld01, n_nonnull_fld02, etc.
with a stats description, to indicate how many non-null values
exist for the nth field of the index. This value is retrieved
and stored in the index statistics
in dict_stats_fetch_index_stats_step(). The counts are per
column, not per n-column prefix.

rec_get_n_blob_pages(): Calculate the number of
externally stored pages for a record, using ceiling division
by the usable BLOB page payload (blob_part_size), which differs
between ROW_FORMAT=COMPRESSED (zip_size minus FIL_PAGE_DATA) and
the other formats (srv_page_size minus the BLOB header and the
page trailer). For ROW_FORMAT=COMPRESSED the length in
the field reference is the uncompressed length, so the result is an
upper bound.

When the leaf level is scanned in full, the number of leaf pages that
were scanned is reported as n_leaf_pages for a multi level index.
Before, result.n_leaf_pages was overwritten with
index->stat_n_leaf_pages, which dict_stats_empty_index() had
just set to 1, so every index that took the full scan path
reported n_leaf_pages=1. Single page indexes report 1.
This changes cardinality estimates and
therefore leads to multiple changes in existing test cases.

Non-null values are counted only at the leaf level, since only leaf
pages hold actual records. A full scan of the leaf level counts them
exactly. When the level is sampled, the per column count is derived
from the sampled leaves with the same formula as n_diff:

  n_ordinary_leaf_pages * n_non_null_all_analyzed_pages
                        / n_leaf_pages_to_analyze

This is an estimate for NOT NULL columns as well: the sampled leaves
may hold fewer or more records than the average, and a dive that
stops at a boring page contributes nothing to the sum while still
counting in the divisor.

innodb_rec_per_key(): stat_n_non_null_key_vals[i] holds the
number of records in which the i-th indexed column alone is
not NULL, while what has to be excluded here is the number
of records whose first i+1 columns are all not NULL,
because that is the population which the n-column
prefix statistic stat_n_diff_key_vals[i] has to be
corrected against when innodb_stats_method=nulls_ignored:
with NULLs compared as unequal, every record carrying a NULL
anywhere in the prefix adds a distinct value of its own to n_diff.

PageStats::scan(): n_non_null is accumulated and assigned only
for leaf pages, so that a non-leaf scan cannot leave a node
pointer count behind when scan_below() stops at a boring page
without reaching a leaf.

IndexLevelStats::reset_for_level() also clears n_diff[], and
dict_stats_analyze_index() zero initializes the buffer backing it, so
that a level scan which finds no records (a failed
btr_pcur_open_level(), or a non-leaf page whose first record is not
marked as the leftmost one on the level) leaves n_diff[] at 0 instead
of stale values.

IndexLevelStats::sample_leaf_pages() returns early when the group
boundaries for the prefix are empty, which is the same condition.

IndexLevelStats::analyze_level(): Instead of copying the last record
of the page, retain the latch on the page until the record has been
compared with the first record of the next page

dict_stats_fetch_index_stats_step() no longer resets
stat_n_non_null_key_vals[] while processing an n_diff_pfxNN row:
dict_stats_empty_table() has already cleared the array before the
fetch, and with n_nonnull_fldNN rows now being read too,
that reset would make the result depend on the order in which the
rows arrive.

dict_stats_save(): now static function in dict0stats.cc that
takes the innodb_stats_method value, and is removed from dict0stats.h.
dict_stats_update_persistent() saves the statistics itself, so its
callers no longer have to.

Replaced btr_rec_get_externally_stored_len() with
rec_get_n_blob_pages() in dict0stats.cc.

btr_rec_get_field_ref_offs() and btr_rec_get_field_ref(),
together with the BTR_BLOB_HDR_* macros, were moved from btr0cur.cc
to btr0cur.h so that rec_get_n_blob_pages() can reuse them;

btr_rec_get_field_ref_offs() is now a noexcept function
returning size_t.

Changed stat_n_diff_key_vals and stat_n_non_null_key_vals from
ib_uint64_t* to uint64_t*

len_is_stored(): simplified to a single comparison, which is
equivalent for the unsigned lengths that it is used with.

Removed the unused UNIV_STATS_DEBUG build macro (univ.i) and
turned the DEBUG_PRINTF() helper in dict0stats.cc into an
unconditional no-op
722ff162d854...
722ff162d854de2d4f4aec4631a273a10710fd08
Vladislav Vaintroub
MDEV-40961 fix SBOM generation - randomize UUID

Previously UUID appeared to be constant. Add randomizer to it.
619602e40900...
619602e4090011d151b849b6e9e72315f541a299
Raghunandan Bhat
MDEV-41193: ASAN heap-buffer-overflow in ha_connect::CheckCond after select from Connect table

Problem:
  When CONNECT engine pushes a WHERE clause down to an external table,
  it writes the filter into the work area, without checking how much
  space is left. A large string literal in the WHERE clause can overflow
  the work area allocated by the engine. For ex: if connect_work_size is
  set to 4MB and the string literal in the WHERE clause is larger than
  4MB, it can grow past the allocated work area.

Fix:
  Track the space left in the work area and check it before writing. If
  the filter doesn't fit, drop it instead of writing past the buffer.
5ea07d7b6464...
5ea07d7b64640de9ab65afd0924e5560e2052669
Sergei Golubchik
mariadb_private
5deb85687bce...
5deb85687bce02c244bcaa16ad2bdc9f487e2598
Aleksey Midenkov
MDEV-41050 versioned DELETE via row_end index leaves a row undeleted

1. Aria/MyISAM engines

A system-versioned DELETE on a MyISAM or Aria table indexed by row_end
could leave the last current row undeleted. The delete scans the current
rows via an equality search on row_end = MAX, which is driven by
mi_rnext_same/maria_rnext_same. That function keeps the search's
reference key in lastkey2 and uses the HA_STATE_RNEXT_SAME flag to
remember it has already stored it. This reference-key mechanism is
exactly what lets the scan modify rows it is walking without skipping
them.

Deleting a versioned row is an in-place update of row_end, and the update
reuses lastkey2 as scratch space for the changed key, so it clears
HA_STATE_RNEXT_SAME to request that rnext_same re-store its reference on
the next call. However, TABLE::delete_row wraps the update in
HA_EXTRA_REMEMBER_POS/HA_EXTRA_RESTORE_POS, and RESTORE_POS restored the
whole saved info->update word, resurrecting the HA_STATE_RNEXT_SAME bit
that the update had just cleared.

As a result rnext_same skipped rebuilding its reference key and compared
subsequent keys against the now-overwritten lastkey2, hitting a spurious
end-of-file and terminating the scan one row early, defeating the engine's
own protection against a Halloween-style skip.

Fixed by preserving the current HA_STATE_RNEXT_SAME bit across
RESTORE_POS instead of restoring the stale saved value. See also the HEAP
fix below: same root cause, different per-engine mechanism.

2. HEAP engine

The row loss also reproduces on the MEMORY (HEAP) engine. A system-
versioned DELETE scans the current rows on the row_end index and turns
each delete into an in-place update of row_end, so it modifies the very
index it is walking.

When the changed key is the scanned one (info->lastinx), hp_delete_key()
repositions the cursor but heap_update() leaves info->update untouched, so
HA_STATE_NEXT_FOUND from the preceding heap_rnext() stays set. The next
heap_rnext() then sees current_ptr == 0 with that bit and takes the
"!current_ptr && HA_STATE_NEXT_FOUND" guard as a false end-of-file,
stopping one row early.

Fixed by clearing HA_STATE_NEXT_FOUND when the scanned index key changed.
HA_STATE_AKTIV is kept (unlike heap_delete): the row is updated, not
removed, so a following op must not fail test_active(). The bit is only
set after a heap_rnext(), so a plain single-row UPDATE never reaches this.
See also the Aria/MyISAM fix above: same root cause, different per-engine
mechanism.

3. Why the fix differs per engine, and InnoDB

Aria and HEAP both trace their handler design back to MyISAM, hence the
same root cause (stale scan bookkeeping after an in-place key change) in
all three, fixed at each engine's own bookkeeping spot. InnoDB needs no
fix: its persistent cursor survives concurrent index modification by
design, already covered by this same test under the timestamp
combination (default-storage-engine=innodb), which passes unmodified.
57cf0f62951a...
57cf0f62951a94565ace0ebaf36767a06373e195
Dave Gosselin
MDEV-28509: Dereferenced null pointer of type 'struct JOIN_TAB' in add_key_field

This patch fixes a crash when calculating join statistics during query
optimization for queries with an unused WINDOW definition.  Put another
way, the system may crash when a query defines a WINDOW but doesn't then
refer to it.

Item::marker is overloaded for different uses, many of which treat it as
a bit field.  However, the setup_group function used it to mark that a
field was found when traversing a GROUP BY.  Originally, this marking
set the Item::marker field to 1 to indicate that it was found.  Later
on in setup_group (and only when SQL mode ONLY_FULL_GROUP_BY is
enabled), we would skip any such marked fields when checking that
fields only referenced those found in the GROUP BY; otherwise, it would
be silly to find fields of the GROUP BY within the GROUP BY field
itself.  Setting Item::marker to 1 seemed mostly harmless at that point
in time.  But later, in git sha 4d143a6ff6, we introduced several
changes: (1) the value of marker in setup_group was changed from 1 to
UNDEF_POS, (2) Item::marker was changed from uint8 to int8, and
(3) UNDEF_POS which is defined to be -1 was also added.

Queries that define WINDOWs internally will setup groups and orders as
part of query processing via the setup_group function.  Consequently
because of the behavior described earlier above, such queries may have
items with markers as MARKER_UNDEF_POS (-1) which is the same
as marking all of the flag bits as set.  This is disastrous for those
users of Item::marker which refer to it as a bit field, because every
flag bit appears set at once, including bits that are mutually
exclusive in meaning.  Even a masked test such as marker &
MARKER_SUBSTITUTION returns true when marker is -1.  In particular, the
method Item_direct_view_ref::grouping_field_transformer_for_where then
treats the ref as flagged for substitution and takes the wrong
execution path, leading to the crash.

The only reader of that marker was the GROUP BY membership test in
setup_group itself, so setup_group no longer writes marker.  A
helper, item_in_group_list, yields the same info by walking the
GROUP BY list instead.
54e0db0565c8...
54e0db0565c8e07a9ef6045286472aee1597b5af
Thirunarayanan Balathandayuthapani
MDEV-19574: innodb_stats_method is not honored when innodb_stats_persistent=ON

Problem:
=======
When persistent statistics are enabled (innodb_stats_persistent=ON),
the innodb_stats_method setting is not properly utilized during
statistics calculation.

The statistics collection functions always use a hardcoded default
behavior for NULL value comparison instead of respecting the
configured stats method. This affects the accuracy of
n_diff_key_vals (distinct key count), particularly for
indexes with nullable columns containing NULL values.

Moreover, stat_n_non_null_key_vals[] was never computed for
persistent statistics; it stayed at the 0 that
dict_stats_empty_index() assigns.

With innodb_stats_method=nulls_ignored, innodb_rec_per_key()
therefore always found n_diff <= n_null and reported one record
per key for every index. This impacts the query optimizer,
which makes decisions based on inaccurate cardinality estimates.

Solution:
========
Introduced IndexLevelStats to collect statistics at a specific
B-tree level during index analysis.

Introduced PageStats to collect statistics for leaf page analysis.

Refactored the following functions:
dict_stats_analyze_index_level() to IndexLevelStats::analyze_level()
dict_stats_analyze_index_for_n_prefix() to IndexLevelStats::sample_leaf_pages()
dict_stats_analyze_index_below_cur() to PageStats::scan_below()
dict_stats_scan_page() to PageStats::scan()

The innodb_stats_method value is read once per table in
dict_stats_update_persistent() and passed down, so that all
indexes of a table are analyzed with the same method.

Add the stats method name to stat_description when
innodb_stats_method has a non-default value. The suffix is
dropped when the description is already full.

Added the new stat name n_nonnull_fld01, n_nonnull_fld02, etc.
with a stats description, to indicate how many non-null values
exist for the nth field of the index. This value is retrieved
and stored in the index statistics
in dict_stats_fetch_index_stats_step(). The counts are per
column, not per n-column prefix.

rec_get_n_blob_pages(): Calculate the number of
externally stored pages for a record, using ceiling division
by the usable BLOB page payload (blob_part_size), which differs
between ROW_FORMAT=COMPRESSED (zip_size minus FIL_PAGE_DATA) and
the other formats (srv_page_size minus the BLOB header and the
page trailer). For ROW_FORMAT=COMPRESSED the length in
the field reference is the uncompressed length, so the result is an
upper bound.

When the leaf level is scanned in full, the number of leaf pages that
were scanned is reported as n_leaf_pages for a multi level index.
Before, result.n_leaf_pages was overwritten with
index->stat_n_leaf_pages, which dict_stats_empty_index() had
just set to 1, so every index that took the full scan path
reported n_leaf_pages=1. Single page indexes report 1.
This changes cardinality estimates and
therefore leads to multiple changes in existing test cases.

Non-null values are counted only at the leaf level, since only leaf
pages hold actual records. A full scan of the leaf level counts them
exactly. When the level is sampled, the per column count is derived
from the sampled leaves with the same formula as n_diff:

  n_ordinary_leaf_pages * n_non_null_all_analyzed_pages
                        / n_leaf_pages_to_analyze

This is an estimate for NOT NULL columns as well: the sampled leaves
may hold fewer or more records than the average, and a dive that
stops at a boring page contributes nothing to the sum while still
counting in the divisor.

innodb_rec_per_key(): stat_n_non_null_key_vals[i] holds the
number of records in which the i-th indexed column alone is
not NULL, while what has to be excluded here is the number
of records whose first i+1 columns are all not NULL,
because that is the population which the n-column
prefix statistic stat_n_diff_key_vals[i] has to be
corrected against when innodb_stats_method=nulls_ignored:
with NULLs compared as unequal, every record carrying a NULL
anywhere in the prefix adds a distinct value of its own to n_diff.

PageStats::scan(): n_non_null is accumulated and assigned only
for leaf pages, so that a non-leaf scan cannot leave a node
pointer count behind when scan_below() stops at a boring page
without reaching a leaf.

IndexLevelStats::reset_for_level() also clears n_diff[], and
dict_stats_analyze_index() zero initializes the buffer backing it, so
that a level scan which finds no records (a failed
btr_pcur_open_level(), or a non-leaf page whose first record is not
marked as the leftmost one on the level) leaves n_diff[] at 0 instead
of stale values.

IndexLevelStats::sample_leaf_pages() returns early when the group
boundaries for the prefix are empty, which is the same condition.

IndexLevelStats::analyze_level(): Instead of copying the last record
of the page, retain the latch on the page until the record has been
compared with the first record of the next page

dict_stats_fetch_index_stats_step() no longer resets
stat_n_non_null_key_vals[] while processing an n_diff_pfxNN row:
dict_stats_empty_table() has already cleared the array before the
fetch, and with n_nonnull_fldNN rows now being read too,
that reset would make the result depend on the order in which the
rows arrive.

dict_stats_save(): now static function in dict0stats.cc that
takes the innodb_stats_method value, and is removed from dict0stats.h.
dict_stats_update_persistent() saves the statistics itself, so its
callers no longer have to.

Replaced btr_rec_get_externally_stored_len() with
rec_get_n_blob_pages() in dict0stats.cc.

btr_rec_get_field_ref_offs() and btr_rec_get_field_ref(),
together with the BTR_BLOB_HDR_* macros, were moved from btr0cur.cc
to btr0cur.h so that rec_get_n_blob_pages() can reuse them;

btr_rec_get_field_ref_offs() is now a noexcept function
returning size_t.

Changed stat_n_diff_key_vals and stat_n_non_null_key_vals from
ib_uint64_t* to uint64_t*

len_is_stored(): simplified to a single comparison, which is
equivalent for the unsigned lengths that it is used with.

Removed the unused UNIV_STATS_DEBUG build macro (univ.i) and
turned the DEBUG_PRINTF() helper in dict0stats.cc into an
unconditional no-op
5408ca1eebcc...
5408ca1eebcc4549cb8f0c744302c9cb1a36e3bf
Alessandro Vetere
MDEV-40408 btr_page_reorganize_low() uses the buffer pool just to obtain a scratch block

Add buf_pool.scratch_buf, a pool of page frames that the page
reorganization operations use instead of taking a block from the global
buffer pool.

buf_pool_t::scratch_buffer: A singly linked list of chunks of
buf_tmp_buffer_t slots. A thread holds at most one slot at a time, and
it holds a page latch while doing so, which is why reserve() must not
wait for the buffer pool or for I/O. When every slot is in use,
reserve() appends a chunk that holds twice the slots of the last one, up
to a limit on the size of one chunk. Only grow() waits, on the mutex
that serializes it and on the allocator. A chunk is never moved or freed
before close(), so a thread can keep using the slot that it reserved
while another thread appends a chunk. Debug builds start with a single
slot, so that a second concurrent page reorganization exercises grow().

buf_pool_t::scratch_buffer::shrink(): Free the page frames of the slots
that are not in use. The slots and the chunks are kept, because a slot
costs 32 bytes while a page frame costs srv_page_size. A frame survives
the first pass, because that pass only clears the used flag. The master
thread calls this often while the server is idle and rarely while it is
active, and buf_pool_t::garbage_collect() calls it under memory
pressure, where it ignores the used flag, because releasing these frames
is much cheaper than shrinking the buffer pool.

buf_pool_t::io_buf_t::acquire(): Factor out the scan for an unreserved
slot, which io_buf_t::reserve() ran twice and the scratch buffer reuses.

btr_page_reorganize_low(), page_zip_reorganize(): Obtain the scratch
page frame from buf_pool.scratch_buf. This removes the buf_pool.mutex
acquisition and the free block wait that buf_block_alloc() could
perform while at least a page X-latch was being held.
btr_page_reorganize_low() also used to leak the block on its error
paths; a single exit now releases the slot.

page_copy_rec_list_end_no_locks(), lock_move_reorganize_page(): Take
the source page frame instead of a source buf_block_t, because the
source is no longer a buffer pool block. In
page_copy_rec_list_end_no_locks() the source page is page_align(rec) at
every call site, so only the record is passed.

buf_tmp_buffer_t::acquire(), buf_tmp_buffer_t::release(): Use acquire
and release memory ordering, so that the page frame pointer of a slot
is published to the next thread that reserves the slot. acquire() also
reads the flag before the exchange, so that a scan across an array of
slots does not write to the slots that it finds reserved. release()
also marks the page frame undefined for Valgrind and MSan, because the
frame stays allocated for the next reserver and no deallocation marks
it.
511d7527eb35...
511d7527eb356aab52fa183456333c5ba9741bcf
Dave Gosselin
MDEV-28509:  Dereferenced null pointer of type 'struct JOIN_TAB' in add_key_field

setup_group no longer writes Item::marker.  The ONLY_FULL_GROUP_BY
check now tests GROUP BY membership by walking the GROUP BY list.

A query that defines a WINDOW but never refers to it could crash in
add_key_field, for example

  WITH cte AS (SELECT i FROM (SELECT i FROM t1 GROUP BY i) dt
              WINDOW w AS (PARTITION BY i))
  SELECT a.i FROM cte a JOIN cte b ON a.i=b.i WHERE a.i != 5;

A query that defines a WINDOW goes through setup_group, which set
marker to MARKER_UNDEF_POS (-1) on each GROUP BY expression so that
the ONLY_FULL_GROUP_BY check could skip it.  Other code reads marker
as a set of flag bits (-1 sets all bits).
Item_direct_view_ref::grouping_field_transformer_for_where then took
the ref as flagged for substitution and followed a path that ends in
the crash.

The ONLY_FULL_GROUP_BY check was the only reader of that value, so
MARKER_UNDEF_POS is removed.  The necessary check is local to the
setup_group function.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
4cd714136f9e...
4cd714136f9e0cc75e31890e31758c66ed91b450
Vladislav Vaintroub
MDEV-40608 build mysqlservices without an embedded CRT requirement

mysqlservices only exposes a thin C API, no CRT state crosses it, so
don't force whatever CRT/config built the server onto a plugin linking
it. Without /Zl, a plugin built in a config with no matching installed
mysqlservices variant (CMake silently substitutes one - verified with
a toy project) gets an ignorable but noisy LNK4098 warning.

Assisted-by: Claude:claude-5-sonnet
3815a7ecefd1...
3815a7ecefd19558be2d8abedf871b6f051a8836
Thirunarayanan Balathandayuthapani
MDEV-19574: innodb_stats_method is not honored when innodb_stats_persistent=ON

Problem:
=======
When persistent statistics are enabled (innodb_stats_persistent=ON),
the innodb_stats_method setting is not properly utilized during
statistics calculation.

The statistics collection functions always use a hardcoded default
behavior for NULL value comparison instead of respecting the
configured stats method. This affects the accuracy of
n_diff_key_vals (distinct key count), particularly for
indexes with nullable columns containing NULL values.

Moreover, stat_n_non_null_key_vals[] was never computed for
persistent statistics; it stayed at the 0 that
dict_stats_empty_index() assigns.

With innodb_stats_method=nulls_ignored, innodb_rec_per_key()
therefore always found n_diff <= n_null and reported one record
per key for every index. This impacts the query optimizer,
which makes decisions based on inaccurate cardinality estimates.

Solution:
========
Introduced IndexLevelStats to collect statistics at a specific
B-tree level during index analysis.

Introduced PageStats to collect statistics for leaf page analysis.

Refactored the following functions:
dict_stats_analyze_index_level() to IndexLevelStats::analyze_level()
dict_stats_analyze_index_for_n_prefix() to IndexLevelStats::sample_leaf_pages()
dict_stats_analyze_index_below_cur() to PageStats::scan_below()
dict_stats_scan_page() to PageStats::scan()

The innodb_stats_method value is read once per table in
dict_stats_update_persistent() and passed down, so that all
indexes of a table are analyzed with the same method.

Add the stats method name to stat_description when
innodb_stats_method has a non-default value. The suffix is
dropped when the description is already full.

Added the new stat name n_nonnull_fld01, n_nonnull_fld02, etc.
with a stats description, to indicate how many non-null values
exist for the nth field of the index. This value is retrieved
and stored in the index statistics
in dict_stats_fetch_index_stats_step(). The counts are per
column, not per n-column prefix.

rec_get_n_blob_pages(): Calculate the number of
externally stored pages for a record, using ceiling division
by the usable BLOB page payload (blob_part_size), which differs
between ROW_FORMAT=COMPRESSED (zip_size minus FIL_PAGE_DATA) and
the other formats (srv_page_size minus the BLOB header and the
page trailer). For ROW_FORMAT=COMPRESSED the length in
the field reference is the uncompressed length, so the result is an
upper bound.

When the leaf level is scanned in full, the number of leaf pages that
were scanned is reported as n_leaf_pages for a multi level index.
Before, result.n_leaf_pages was overwritten with
index->stat_n_leaf_pages, which dict_stats_empty_index() had
just set to 1, so every index that took the full scan path
reported n_leaf_pages=1. Single page indexes report 1.
This changes cardinality estimates and
therefore leads to multiple changes in existing test cases.

Non-null values are counted only at the leaf level, since only leaf
pages hold actual records. A full scan of the leaf level counts them
exactly. When the level is sampled, the per column count is derived
from the sampled leaves with the same formula as n_diff:

  n_ordinary_leaf_pages * n_non_null_all_analyzed_pages
                        / n_leaf_pages_to_analyze

This is an estimate for NOT NULL columns as well: the sampled leaves
may hold fewer or more records than the average, and a dive that
stops at a boring page contributes nothing to the sum while still
counting in the divisor.

innodb_rec_per_key(): stat_n_non_null_key_vals[i] holds the
number of records in which the i-th indexed column alone is
not NULL, while what has to be excluded here is the number
of records whose first i+1 columns are all not NULL,
because that is the population which the n-column
prefix statistic stat_n_diff_key_vals[i] has to be
corrected against when innodb_stats_method=nulls_ignored:
with NULLs compared as unequal, every record carrying a NULL
anywhere in the prefix adds a distinct value of its own to n_diff.

PageStats::scan(): n_non_null is accumulated and assigned only
for leaf pages, so that a non-leaf scan cannot leave a node
pointer count behind when scan_below() stops at a boring page
without reaching a leaf.

IndexLevelStats::reset_for_level() also clears n_diff[], and
dict_stats_analyze_index() zero initializes the buffer backing it, so
that a level scan which finds no records (a failed
btr_pcur_open_level(), or a non-leaf page whose first record is not
marked as the leftmost one on the level) leaves n_diff[] at 0 instead
of stale values.

IndexLevelStats::sample_leaf_pages() returns early when the group
boundaries for the prefix are empty, which is the same condition.

IndexLevelStats::analyze_level(): Instead of copying the last record
of the page, retain the latch on the page until the record has been
compared with the first record of the next page

dict_stats_fetch_index_stats_step() no longer resets
stat_n_non_null_key_vals[] while processing an n_diff_pfxNN row:
dict_stats_empty_table() has already cleared the array before the
fetch, and with n_nonnull_fldNN rows now being read too,
that reset would make the result depend on the order in which the
rows arrive.

dict_stats_save(): now static function in dict0stats.cc that
takes the innodb_stats_method value, and is removed from dict0stats.h.
dict_stats_update_persistent() saves the statistics itself, so its
callers no longer have to.

Replaced btr_rec_get_externally_stored_len() with
rec_get_n_blob_pages() in dict0stats.cc.

btr_rec_get_field_ref_offs() and btr_rec_get_field_ref(),
together with the BTR_BLOB_HDR_* macros, were moved from btr0cur.cc
to btr0cur.h so that rec_get_n_blob_pages() can reuse them;

btr_rec_get_field_ref_offs() is now a noexcept function
returning size_t.

Changed stat_n_diff_key_vals and stat_n_non_null_key_vals from
ib_uint64_t* to uint64_t*

len_is_stored(): simplified to a single comparison, which is
equivalent for the unsigned lengths that it is used with.

Removed the unused UNIV_STATS_DEBUG build macro (univ.i) and
turned the DEBUG_PRINTF() helper in dict0stats.cc into an
unconditional no-op
2d346ffb1743...
2d346ffb174372317b28d6088219c9e2f7534ff5
Aleksey Midenkov
MDEV-41050 versioned DELETE via row_end index leaves a row undeleted

1. Aria/MyISAM engines

A system-versioned DELETE on a MyISAM or Aria table indexed by row_end
could leave the last current row undeleted. The delete scans the current
rows via an equality search on row_end = MAX, which is driven by
mi_rnext_same/maria_rnext_same. That function keeps the search's
reference key in lastkey2 and uses the HA_STATE_RNEXT_SAME flag to
remember it has already stored it. This reference-key mechanism is
exactly what lets the scan modify rows it is walking without skipping
them.

Deleting a versioned row is an in-place update of row_end, and the update
reuses lastkey2 as scratch space for the changed key, so it clears
HA_STATE_RNEXT_SAME to request that rnext_same re-store its reference on
the next call. However, TABLE::delete_row wraps the update in
HA_EXTRA_REMEMBER_POS/HA_EXTRA_RESTORE_POS, and RESTORE_POS restored the
whole saved info->update word, resurrecting the HA_STATE_RNEXT_SAME bit
that the update had just cleared.

As a result rnext_same skipped rebuilding its reference key and compared
subsequent keys against the now-overwritten lastkey2, hitting a spurious
end-of-file and terminating the scan one row early, defeating the engine's
own protection against a Halloween-style skip.

Fixed by preserving the current HA_STATE_RNEXT_SAME bit across
RESTORE_POS instead of restoring the stale saved value. See also the HEAP
fix below: same root cause, different per-engine mechanism.

2. HEAP engine

The row loss also reproduces on the MEMORY (HEAP) engine. A system-
versioned DELETE scans the current rows on the row_end index and turns
each delete into an in-place update of row_end, so it modifies the very
index it is walking.

When the changed key is the scanned one (info->lastinx), hp_delete_key()
repositions the cursor but heap_update() leaves info->update untouched, so
HA_STATE_NEXT_FOUND from the preceding heap_rnext() stays set. The next
heap_rnext() then sees current_ptr == 0 with that bit and takes the
"!current_ptr && HA_STATE_NEXT_FOUND" guard as a false end-of-file,
stopping one row early.

Fixed by clearing HA_STATE_NEXT_FOUND when the scanned index key changed.
HA_STATE_AKTIV is kept (unlike heap_delete): the row is updated, not
removed, so a following op must not fail test_active(). The bit is only
set after a heap_rnext(), so a plain single-row UPDATE never reaches this.
See also the Aria/MyISAM fix above: same root cause, different per-engine
mechanism.

3. Why the fix differs per engine, and InnoDB

Aria and HEAP both trace their handler design back to MyISAM, hence the
same root cause (stale scan bookkeeping after an in-place key change) in
all three, fixed at each engine's own bookkeeping spot. InnoDB needs no
fix: its persistent cursor survives concurrent index modification by
design, already covered by this same test under the timestamp
combination (default-storage-engine=innodb), which passes unmodified.
249c1290852d...
249c1290852d0604cc14ce67cc75d55b79a50f46
Georg Richter
CONC-853: Fix TLS verification error mapping parity between OpenSSL and GnuTLS

Align certificate verification behavior and status bitmask propagation
across OpenSSL and GnuTLS backends. Previously, verification discrepancies
led to false negatives, silent auth fallbacks, and assertion failures in
`my_auth.c` due to inconsistent mapping of self-signed vs. untrusted leaf
certificates and improper bitmask evaluation.

Partial rewrite of tls_server.py
* Replace deprecated `pyOpenSSL` dependency with `cryptography` library
  for dynamic SSL context and test certificate generation in tls_server.py.
* Ensure full parity and coverage for `test_tls_verify_unknown` across both
  OpenSSL and GnuTLS backend builds.
  • cc-x-codbc-windows: 'dojob pwd if '3.4' == '3.4' ls win32/test SET TEST_DSN=master SET TEST_DRIVER=master SET TEST_PORT=3306 SET TEST_SCHEMA=odbcmaster if '3.4' == '3.4' cd win32/test if '3.4' == '3.4' ctest --output-on-failure' failed -  stdio
15c9a2207198...
15c9a2207198d0715564d4502eaab92b9119de35
Vladislav Vaintroub
MDEV-40967 PROXY protocol host check sent in clear text mid-SSL handshake

Defer the host-privileged/host-blocked check for a PROXY-header-derived
address until after the client's SSL handshake completes, instead of
sending it immediately in clear text.

Co-Authored-By: Claude Sonnet 5 <[email protected]>
02a56c4d7d22...
02a56c4d7d22678ba826c5c3a5e4849b05a7ae4f
Marko Mäkelä
WIP MDEV-41010 BACKUP SERVER TO is slower than mariadb-backup

mariadb-backup --backup always uses a dedicated log_copying_thread()
that eagerly copies the log from the server. Let us do the same
in multi-threaded BACKUP SERVER TO.

FIXME: revise the HAVE_INNODB_PMEM code path and remove log_mutex.

backup_sink::id: The thread identifier (0 to CONCURRENT-1)

InnoDB_backup::context::tracked: Queue of log files to be copied.

innodb_backup_checkpoint_pmem(): Copy the old log file.

InnoDB_backup::log_track(), InnoDB_backup::log_track_pmem():
Keep copying the log until we run out of InnoDB data files to copy.

InnoDB_backup::checkpoint_complete_pmem(): Copy the remaining
part of an old log file right before it is being released.

InnoDB_backup::commit(): In log tracking backup, copy the rest of
the HAVE_INNODB_PMEM log. (FIXME: buffer it for log_track_pmem())
BuildBot (0.8.9) working for the MariaDB project.
Page built: Wed 23 Sep 2026 15:28:27 (UTC)