alexpeshkoff
7fed9ff78c
cleanup
2015-03-05 16:18:46 +00:00
hvlad
39a137e913
Use attachment's or relation's pool instead of dbb_permanent where it is more suitable.
...
It fixed memory leak from dbb_permanent (CORE-4683)
2015-02-22 11:02:49 +00:00
dimitr
c6f6557d5e
Misc.
2015-02-19 15:53:42 +00:00
dimitr
6ae3e2cbaf
1) Fixed CORE-4383: Index and BLOBs garbage collection doesn't work for update_in_place().
...
2) Applied the same approach to temporary GC records.
3) Refactored the Record class to avoid reallocations, to improve encapsulation and to simplify the code.
4) Slightly refactored UndoItem to isolate its logic from the Record class.
5) Reduced the in-memory size of the undo log.
6) Slightly better alternative to the legacy NULL reporting for outer joins.
7) Minor fixes and adjustments in BufferedStream.
2015-02-19 14:15:00 +00:00
hvlad
79b760b4ac
Memory leak (CORE-4683)
2015-02-16 13:57:28 +00:00
dimitr
c8d6349c77
Replace a verb-cleanup bugcheck with regular error + transaction invalidation.
2015-02-03 16:31:05 +00:00
asfernandes
a04966874e
Correction.
2015-01-12 00:21:07 +00:00
asfernandes
c4228eb623
Misc.
2015-01-12 00:04:14 +00:00
starodubov
b260c1ead1
Fixed CORE-4656: Server could hang during chasing dead record version
2014-12-30 08:17:19 +00:00
dimitr
0e11867e84
Refactored the RPB flags. The new logic also provides a second-level protection against CORE-4618.
2014-12-01 21:53:29 +00:00
dimitr
9b0a24fc5a
Simplest possible fix for CORE-4618: Rollback doesn`t undo changes when MERGE statement updates the same target rows multiple times and PLAN MERGE is used. An alternative solution is being considered.
2014-11-30 16:00:19 +00:00
dimitr
d98625b933
Relaxed system tables protection for cases UPDATE ... SET RDB$*_SOURCE = NULL. So far nullification is allowed for views/packages/procedures/functions/triggers only. I'm not sure somebody really needs hiding sources for defaults and check constraints, but this can be supported later, if required. This is a temporary solution, intended to be replaced with something more clever (e.g. special permissions or new DDL command) in v4.
2014-11-09 16:40:52 +00:00
hvlad
18806682fd
Fixed bug CORE-4575 : AV in garbage collector thread at disconnect of last attachment
2014-10-10 10:11:14 +00:00
dimitr
9defd49c6d
Count record reads performed by the IDX code.
2014-09-01 08:29:04 +00:00
dimitr
6d77447723
Minor refactoring.
2014-08-28 15:53:43 +00:00
dimitr
82bf6c2c0e
Minor adjustments to the performance counters. Work in progress.
2014-08-28 15:16:39 +00:00
dimitr
4dae12ccce
Added counter for record refetches (aka repeated reads).
2014-08-27 14:51:51 +00:00
dimitr
a79ff180da
Added specific error code for read conflicts in RC NO_REC_VER mode. Added missing conflict counting.
2014-08-27 12:56:36 +00:00
dimitr
36af3d8cb2
Cleanup and small refactoring to avoid redundancy.
2014-08-15 14:19:02 +00:00
dimitr
abb5897fce
Fixed possible deadlocks between a worker thread and an attachment-level AST.
2014-08-08 15:27:59 +00:00
hvlad
79637fc733
Missing notification of gc thread.
2014-08-06 08:55:47 +00:00
dimitr
a8b1dba646
Refactored RuntimeStatistics internals. Added more record-level performance counters.
2014-08-05 08:46:09 +00:00
asfernandes
718eb6a33e
Fixed the restore of packaged routines.
2014-08-01 16:22:37 +00:00
dimitr
6199136792
Refactored database monitoring, made its handling attachment bound. This should make deadlocks impossible, as well as avoid blocking worker connections while the monitoring request is processed.
2014-07-31 08:56:53 +00:00
hvlad
8741e3d68c
Cleanup and optimize locking a bit (as suggested by Dmitry)
2014-07-29 08:15:29 +00:00
skidder
e2e901939c
Fix GCC warnings
2014-07-17 21:17:15 +00:00
skidder
7c743fc291
Fix build warnings
2014-07-17 18:48:46 +00:00
roman-simakov
d74aeda1a0
Corrections pointed by Adriano.
2014-07-13 18:50:57 +00:00
roman-simakov
573e93e5a6
Fixed CORE-735: User rights for metadata changes.
...
We check DDL in DDL nodes and skip at vio level. vio level still exists for direct metadata editing.
2014-07-08 07:35:27 +00:00
alexpeshkoff
26319333e3
Renamed mapping-related tables as was discussed in devel
2014-06-26 09:37:55 +00:00
dimitr
153b06b2fd
Minor structure changes.
2014-06-13 16:28:22 +00:00
alexpeshkoff
6a2016235a
Fixed CORE-4435: After calling release() for attachment to database (instead detach) in embedded mode attachment remains active forever (and some cleanup re. reference counters)
2014-06-10 07:13:34 +00:00
asfernandes
3f8de25fcc
Feature CORE-4403 - Allow referencing cursors as record variables in PSQL.
...
Improvement CORE-4434 - Extend the use of colon prefix for read/assignment OLD/NEW fields and assignment to variables.
2014-05-19 19:06:47 +00:00
robocop
492f4437c4
Misplaced parentheses are unexpected invitees.
2014-05-10 14:17:53 +00:00
robocop
632075d98f
It seems it's possible to restrict gbak more on deletions. Thanks Vlad for hint on TDBB_dont_post_dfw.
2014-05-09 20:26:01 +00:00
asfernandes
40b38014e4
Misc.
2014-05-08 23:26:40 +00:00
alexpeshkoff
bf82df1ebb
Postfix for CORE-3242: somewhy when checking metadata access (like MODIFY or DROP) always checked that access from object itself. Strange at the first glance behavior, but let us do not change legacy (at least since FB1.0) when possible.
2014-05-07 11:21:31 +00:00
robocop
4895ac6067
Fix the problem with gbak restoring a db and lack of rights on sys tables that Philippe detected.
2014-05-07 08:58:36 +00:00
robocop
e6d3b34c55
Fix CORE-4418, using the property of special triggers that are marked "system" or "ignore_perm".
2014-05-06 01:28:53 +00:00
robocop
a6fc5ea73b
1.- Don't let gbak delete more than the only table it might clean. 2.- Special case for rdb$types: user can insert records, provided that's locksmith and the records contain non-system types.
2014-05-02 03:01:38 +00:00
robocop
8c83e71e88
More sys tables protected.
2014-05-01 05:55:15 +00:00
dimitr
a6225eca2d
Attempted to fix CORE-4388: SELECT WITH LOCK may enter an infinite loop for a single record. To be field tested.
2014-04-13 08:27:04 +00:00
robocop
6fcd20e73f
Experimental way of stopping gbak impersonation in sys tables. Gbak only needs to write to sys tables when it's restoring and only in the first attachment. Second and third attachment are for bringing the DB online and setting the R/O flag. Please test.
2014-04-10 01:39:32 +00:00
robocop
de7ed79414
Fixing several issues with my previous commit:
...
Use increment [by], no step.
Enable again code in ini.epp because it's needed when restoring (but not when creating gens directly in a new db).
Gbak has code to fix system generators and we should allow it.
Nbackup cannot update the history generator when doing a backup.
Field name renamed to rdb$generator_increment.
2014-04-09 07:17:34 +00:00
robocop
29e36dc4f3
The 0th generator becomes unnamed. Only internal requests can access it (zero length names are not allowed from outside). RDB$GENERATORS is no longer a reserved name.System generators cannot be changed by user requests, but can be queried with gen_id(g, 0) if the permissions (ACL) allow that. We don't track dependencies on sys generators, it's not necessary.
2014-04-04 06:39:09 +00:00
robocop
315a1b79fc
Misc.
2014-03-22 20:56:13 +00:00
asfernandes
f197d5912a
Fixed restore of sh_test.gbk in TCS.
2014-03-14 16:45:09 +00:00
robocop
e964c4f2a2
More restrictions on sys tables. Please report if you have problems with DDL or the utilities.
2014-03-11 07:40:42 +00:00
robocop
aac88ea33d
Make BTR_lookup return bool.
2014-02-25 06:30:18 +00:00
hvlad
2a67db1b06
Better (i hope) fix for the bugcheck appeared in test bugs.core_195.
...
The original fix broke cursor stability when there are sub-queries, see also CORE-3362.
Misc fixes for better debugging.
2014-01-09 10:29:37 +00:00
