]>
git.sur5r.net Git - openldap/log
Howard Chu [Fri, 7 Oct 2011 01:53:07 +0000 (18:53 -0700)]
Fix page_split with MDB_RESERVE
Howard Chu [Thu, 6 Oct 2011 23:11:25 +0000 (16:11 -0700)]
Fix subcursor initialization
Howard Chu [Tue, 4 Oct 2011 01:02:17 +0000 (18:02 -0700)]
Don't use Append mode when splits are caused by sub-page inserts
Sub-pages need room to grow...
Howard Chu [Mon, 3 Oct 2011 19:40:53 +0000 (12:40 -0700)]
Fix freelist bug on page split
NULL out env->pghead before writing it, so that its pages can't be used
to satisfy the write.
Howard Chu [Mon, 3 Oct 2011 18:25:31 +0000 (11:25 -0700)]
Add rudimentary MDB_MULTIPLE support for puts
Only for DUPFIXED DBs. Still could do better.
Howard Chu [Mon, 3 Oct 2011 04:27:07 +0000 (21:27 -0700)]
Add MDB_APPEND to avoid splitting full pages
Howard Chu [Mon, 3 Oct 2011 02:52:12 +0000 (19:52 -0700)]
Howard Chu [Mon, 3 Oct 2011 02:27:31 +0000 (19:27 -0700)]
Add mdb_cursor_txn() and mdb_cursor_dbi() for querying the cursor
Howard Chu [Sun, 2 Oct 2011 17:02:30 +0000 (10:02 -0700)]
Add MDB_RESERVE mode
When putting a record, just make space for the data, don't copy it.
(Not compatible with MDB_DUPSORT, since the actual data is needed
to determine the insert location.)
Howard Chu [Sat, 1 Oct 2011 05:15:05 +0000 (22:15 -0700)]
Tweak cursor adjustment code
Skip the original cursor that made the change
Howard Chu [Fri, 30 Sep 2011 04:07:07 +0000 (21:07 -0700)]
Make sure mdb_open actually got a sub-db record
Howard Chu [Fri, 23 Sep 2011 04:46:12 +0000 (21:46 -0700)]
Doc for the MDB_NOSUBDIR env option
Howard Chu [Fri, 23 Sep 2011 01:05:06 +0000 (18:05 -0700)]
Add MDB_NOSUBDIR option
Create the data file using exactly the given pathname, and
the lock file using a suffix on the data file name.
Howard Chu [Thu, 22 Sep 2011 11:34:40 +0000 (04:34 -0700)]
Reader table notes
Howard Chu [Thu, 22 Sep 2011 00:21:48 +0000 (17:21 -0700)]
assert cleanup
Howard Chu [Wed, 21 Sep 2011 23:37:29 +0000 (16:37 -0700)]
txn_abort/commit free all their cursors now.
Howard Chu [Wed, 21 Sep 2011 23:06:52 +0000 (16:06 -0700)]
Only free malloc'd cursors
Howard Chu [Wed, 21 Sep 2011 21:14:11 +0000 (14:14 -0700)]
Track changes to all cursors.
For any change to a page or node, update all other cursors pointing
at the same page (or node). Cursors are now stored in a linked list
off their owning transaction. Cursors are all closed when the transaction
ends. Cursors in parent transactions are updated when their child
transaction commits.
Howard Chu [Tue, 20 Sep 2011 19:57:05 +0000 (12:57 -0700)]
Add support for nested transactions.
Hallvard Furuseth [Tue, 13 Sep 2011 07:39:28 +0000 (09:39 +0200)]
libmdb IDL cleanup.
Microoptimize IDL search.
Use RANGE_<FIRST/LAST> when IDL is known to be a range.
Howard Chu [Mon, 19 Sep 2011 01:26:54 +0000 (18:26 -0700)]
Additional fix for prev commit
Howard Chu [Mon, 19 Sep 2011 01:13:23 +0000 (18:13 -0700)]
Make sure 2 subpages fit in one real page
Howard Chu [Mon, 19 Sep 2011 00:07:47 +0000 (17:07 -0700)]
Fix last delete from a subpage
Howard Chu [Sun, 18 Sep 2011 23:39:18 +0000 (16:39 -0700)]
Fix
09006ccec7928c9cf53bca6abe741e8d4d466c98
Check for stale DBs was in the wrong place.
Howard Chu [Sun, 18 Sep 2011 21:18:27 +0000 (14:18 -0700)]
Fix mdb_drop() if not deleting the DB record
Howard Chu [Sun, 18 Sep 2011 20:28:08 +0000 (13:28 -0700)]
Fix typo, minor cleanup
Howard Chu [Sun, 18 Sep 2011 19:37:02 +0000 (12:37 -0700)]
Add mdb_drop() for deleting a DB
Allow MIDLs to grow arbitrarily, to allow unlimited number of
pages to be freed in one txn.
Howard Chu [Sun, 18 Sep 2011 17:38:08 +0000 (10:38 -0700)]
API change, consistency check
mdb_close() takes an env, not a txn.
Also, add check for stale DB records (in case some other process
wrote to the DB last).
Howard Chu [Fri, 16 Sep 2011 23:51:32 +0000 (16:51 -0700)]
New sorted-dup subpage support
Instead of converting directly to a subDB when the first duplicate
item is seen for a key, convert to a subpage instead. Allow the
subpage to grow up to the overflow limit, then convert to a subDB.
This saves a significant amount of space in a typical slapd index
database.
Currently we don't convert back to the smaller form if items are
later deleted. Probably could do that with some hysteresis, e.g.,
convert back from subDB to subpage when the size drops below
(overflow limit/2). Maybe later.
Howard Chu [Thu, 15 Sep 2011 17:36:15 +0000 (10:36 -0700)]
Use FNV-1A hash to generate unique mutex names
For Windows and MacOSX
Howard Chu [Thu, 15 Sep 2011 03:33:09 +0000 (20:33 -0700)]
Doc tweak - txn flags
Howard Chu [Thu, 15 Sep 2011 03:30:26 +0000 (20:30 -0700)]
MacOSX semaphore names are limited to 31 chars
Including the terminating NUL. Good job documenting that, guys.
Howard Chu [Wed, 14 Sep 2011 18:31:27 +0000 (11:31 -0700)]
Add MacOSX support
mmap() with FIXEDMAP fails, otherwise things work.
Howard Chu [Wed, 14 Sep 2011 08:46:08 +0000 (01:46 -0700)]
Minor doc cleanup
Howard Chu [Wed, 14 Sep 2011 08:08:01 +0000 (01:08 -0700)]
More docs
Howard Chu [Wed, 14 Sep 2011 06:36:41 +0000 (23:36 -0700)]
Namespace cleanup
Howard Chu [Tue, 13 Sep 2011 20:34:22 +0000 (13:34 -0700)]
Tweak search_node inner loop to avoid LEAF2 checks
Howard Chu [Tue, 13 Sep 2011 18:25:17 +0000 (11:25 -0700)]
cursor_set/search_node tweaks, use intcmp when possible
Howard Chu [Mon, 12 Sep 2011 08:11:56 +0000 (01:11 -0700)]
Minor tweaks, update relfunc behavior
Howard Chu [Mon, 12 Sep 2011 06:57:17 +0000 (23:57 -0700)]
More cursor_init cleanup
Howard Chu [Mon, 12 Sep 2011 06:13:36 +0000 (23:13 -0700)]
One more sub-cursor fix
Howard Chu [Mon, 12 Sep 2011 05:23:35 +0000 (22:23 -0700)]
Merge branch 'mdb.master' of ssh://git-master.openldap.org/~git/git/openldap into mdb.master
Howard Chu [Mon, 12 Sep 2011 03:54:32 +0000 (20:54 -0700)]
Fix
2e3bc39fa94f21d692d8e94183f57aef9122c487
Various mistakes when converting from previous data structures.
Add a few more debug asserts/sanity checks.
Split all "if (foo) return" constructs to separate lines to allow
easier breakpoint setting.
Add mtest6 for checking mdb_split() behavior. This needs to be
expanded to check rebalance/merge cases too.
Hallvard Furuseth [Sun, 11 Sep 2011 11:00:52 +0000 (13:00 +0200)]
Fix mdb debug formats
Hallvard Furuseth [Sun, 11 Sep 2011 09:08:40 +0000 (11:08 +0200)]
Restore missing LAZY_LOCKS semicolon.
Lost in C90 compat commit
c5dad7a6d0eb1b3597beda18b38fe92820df0d30 .
Hallvard Furuseth [Sat, 10 Sep 2011 08:11:56 +0000 (10:11 +0200)]
mdb integer type cleanup: Use MDB_dbi consistently.
This changes the prototype of mdb_env_set_maxdbs().
Hallvard Furuseth [Sat, 10 Sep 2011 18:01:07 +0000 (20:01 +0200)]
Drop ULONG in favor of size_t.
Hallvard Furuseth [Sat, 10 Sep 2011 08:11:56 +0000 (10:11 +0200)]
Add mdb.c:txnid_t for readability.
Maybe later we'll test txnid wraparound with it defined to ushort.
Hallvard Furuseth [Sat, 10 Sep 2011 08:11:56 +0000 (10:11 +0200)]
mdb type cleanup: Consistent reader count types.
Use unsigned int for reader counts/max limits, that's the smallest
change. Don't need uint32_t when mdb_env_set_maxreaders() takes less.
Change prototypes of mdb_env_set_maxreaders,mdb_env_get_maxreaders().
Check the mdb_env_set_maxreaders() argument.
Hallvard Furuseth [Sat, 10 Sep 2011 08:11:55 +0000 (10:11 +0200)]
explain mdl_midl_sort() istack size
Hallvard Furuseth [Sat, 10 Sep 2011 08:11:55 +0000 (10:11 +0200)]
Fix memn(r)cmp of key sizes > INT_MAX. Simplify.
Hallvard Furuseth [Sat, 10 Sep 2011 08:11:55 +0000 (10:11 +0200)]
Check some mdb integer type assumptions
Hallvard Furuseth [Sat, 10 Sep 2011 08:11:55 +0000 (10:11 +0200)]
Put MDB_node.<mn_lo,mn_hi> in host byte order.
Hallvard Furuseth [Sat, 10 Sep 2011 08:11:55 +0000 (10:11 +0200)]
C90 compatibility cleanup in mdb.
Fix void* pointer arithmetic in cintcmp().
Fix invalid ';'s in declarations after possibly-empty macros.
Hallvard Furuseth [Sat, 10 Sep 2011 08:11:55 +0000 (10:11 +0200)]
Cleanup mdb macros.
Parenthesize. Rename GetPageSize -> GET_PAGESIZE since it does not
behave like a function (it has a non-pointer output argument).
Hallvard Furuseth [Sat, 10 Sep 2011 08:11:55 +0000 (10:11 +0200)]
libmdb/.gitignore += test[45], Doxygen output
Howard Chu [Sat, 10 Sep 2011 04:43:22 +0000 (21:43 -0700)]
If put(NOOVERWRITE) exists, return existing data
Howard Chu [Fri, 9 Sep 2011 10:52:12 +0000 (03:52 -0700)]
Minor typos, additional clarification
Howard Chu [Fri, 9 Sep 2011 03:18:29 +0000 (20:18 -0700)]
Partial revert, fix MDB_page definition
Howard Chu [Fri, 9 Sep 2011 00:13:51 +0000 (17:13 -0700)]
More docs
Howard Chu [Thu, 8 Sep 2011 20:11:33 +0000 (13:11 -0700)]
More docs
Howard Chu [Thu, 8 Sep 2011 20:10:17 +0000 (13:10 -0700)]
Speedup cintcmp
Now that we know we're always 2-byte aligned...
Howard Chu [Thu, 8 Sep 2011 17:08:21 +0000 (10:08 -0700)]
Fix previous commit
Howard Chu [Thu, 8 Sep 2011 06:57:50 +0000 (23:57 -0700)]
Merge branch 'mdb.cursor' into mdb.master
Howard Chu [Thu, 8 Sep 2011 06:10:19 +0000 (23:10 -0700)]
Forgot to commit this
Howard Chu [Thu, 8 Sep 2011 06:10:19 +0000 (23:10 -0700)]
Forgot to commit this
Howard Chu [Wed, 7 Sep 2011 09:04:40 +0000 (02:04 -0700)]
Further cursor-based restructuring
Get rid of all of the redundant MDB_*page structs and other cruft.
Howard Chu [Wed, 7 Sep 2011 06:37:37 +0000 (23:37 -0700)]
Cleanup unused code
Howard Chu [Wed, 7 Sep 2011 06:16:45 +0000 (23:16 -0700)]
Fix mdb_move_node again
Howard Chu [Tue, 6 Sep 2011 22:22:53 +0000 (15:22 -0700)]
Fix search_node, add cintcmp
Howard Chu [Tue, 6 Sep 2011 19:54:09 +0000 (12:54 -0700)]
Guarantee nodes are 2-byte aligned
Also fix xcursor_init1 md_dirty flag
Also another mdb_cmp refactoring fix
Howard Chu [Tue, 6 Sep 2011 12:17:24 +0000 (05:17 -0700)]
Change MDB_node to depend on 2-byte alignment
Howard Chu [Mon, 5 Sep 2011 23:54:39 +0000 (16:54 -0700)]
Fix for mdb_cmp refactoring
Howard Chu [Mon, 5 Sep 2011 21:22:46 +0000 (14:22 -0700)]
Minor cleanup
Howard Chu [Mon, 5 Sep 2011 09:36:24 +0000 (02:36 -0700)]
Use mdb_midl_append/sort instead of _insert
Howard Chu [Mon, 5 Sep 2011 07:46:32 +0000 (00:46 -0700)]
mdb_cmp refactoring
Howard Chu [Mon, 5 Sep 2011 01:34:03 +0000 (18:34 -0700)]
Cursor stuff working, no major perf gain
Howard Chu [Sun, 4 Sep 2011 18:59:32 +0000 (11:59 -0700)]
Cursors in-progress, broken
Howard Chu [Sun, 4 Sep 2011 03:42:46 +0000 (20:42 -0700)]
Use cursor->mc_flags
Howard Chu [Sun, 4 Sep 2011 00:24:39 +0000 (17:24 -0700)]
Begin reimplementing basic ops as cursor ops
Howard Chu [Sat, 3 Sep 2011 22:19:24 +0000 (15:19 -0700)]
Add MDB_{FIRST,LAST}_DUP for cursor_get
Howard Chu [Sun, 4 Sep 2011 03:11:07 +0000 (20:11 -0700)]
Fix mdb_split, fix MDB_GET_BOTH
Howard Chu [Fri, 2 Sep 2011 10:09:05 +0000 (03:09 -0700)]
Cleanup midl, start documenting internals
Howard Chu [Fri, 2 Sep 2011 08:09:11 +0000 (01:09 -0700)]
Tweak docs
Howard Chu [Thu, 1 Sep 2011 10:38:48 +0000 (03:38 -0700)]
Ugly-as-sin initial Windows port
Howard Chu [Wed, 31 Aug 2011 11:58:40 +0000 (04:58 -0700)]
API cleanup, doxygen docs for public API
Howard Chu [Mon, 29 Aug 2011 23:55:41 +0000 (16:55 -0700)]
Cleanup, fix freelist alloc
Don't allow new pages for free list to come from the free list.
Otherwise a nasty data self-reference occurs that is too much
trouble to unwind.
Howard Chu [Mon, 29 Aug 2011 10:45:13 +0000 (03:45 -0700)]
Tweak locks, fix race conditions
Howard Chu [Sun, 28 Aug 2011 00:32:24 +0000 (17:32 -0700)]
Reduce excessive memcpy in mdb_split
Howard Chu [Sat, 27 Aug 2011 23:53:58 +0000 (16:53 -0700)]
Fix mdb_split, tweak split_indx if space is tight
Howard Chu [Fri, 26 Aug 2011 20:37:02 +0000 (13:37 -0700)]
Fix update of big data items
Howard Chu [Fri, 26 Aug 2011 07:01:16 +0000 (00:01 -0700)]
Fix free page reclaim and dbs table update
Can re-use free pages 1 txn earlier
Must update all dbs tables on txn commit
Howard Chu [Fri, 26 Aug 2011 02:45:08 +0000 (19:45 -0700)]
More debug, fix MDB_DEL_DUP
Howard Chu [Thu, 25 Aug 2011 09:08:56 +0000 (02:08 -0700)]
More debug output
Howard Chu [Wed, 24 Aug 2011 08:45:42 +0000 (01:45 -0700)]
Add txn_reset / txn_renew
Howard Chu [Mon, 22 Aug 2011 00:49:54 +0000 (17:49 -0700)]
First cut at DUPFIXED support
Also in cursor_set, check the current page (if already set from
before) before starting over again from the root.
Howard Chu [Sat, 20 Aug 2011 00:20:56 +0000 (17:20 -0700)]
Set MINKEYS to 2
For better space utilization
Howard Chu [Fri, 19 Aug 2011 23:43:57 +0000 (16:43 -0700)]
Keep re-using dpages instead of freeing them
Hallvard B Furuseth [Fri, 19 Aug 2011 06:45:57 +0000 (08:45 +0200)]
Avoid cast-away-const for errstr[]
Hallvard B Furuseth [Fri, 19 Aug 2011 18:35:25 +0000 (20:35 +0200)]
Wrap O_DSYNC in MDB_DSYNC.
If O_DSYNC is not defined, that may be due to poor compiler options.
We could fall back to the less efficient O_SYNC, but it seems better
to let the user learn of the problem and give better compiler options.