dimitr
5e9a5ef312
Attempted to fix a number of issues with badly working privileges after restore into ODS12.
2013-11-18 16:28:24 +00:00
alexpeshkoff
82f1524800
Fixed some more warnings
2013-11-18 14:56:14 +00:00
alexpeshkoff
120b10a78a
Implemented CORE-4263: Database linger
2013-11-14 16:16:24 +00:00
asfernandes
dc99961e8b
Misc.
2013-10-13 23:44:57 +00:00
hvlad
1267bfb3f6
Fix to please GCC
2013-09-13 08:15:17 +00:00
hvlad
db70a7742d
Fixed bug CORE-3921 : "record disappeared (186), file: vio.cpp line: 408"
2013-09-12 19:08:59 +00:00
hvlad
cf4e7ecc0d
Fixed bug CORE-4212 : Dropping FK on GTT crashes server
2013-09-10 14:30:42 +00:00
dimitr
8586171695
Core support for records longer than 64KB. Not exposed to the external world yet.
2013-08-21 07:40:31 +00:00
dimitr
69961e9bae
Fixed CORE-4164: Owner name is missing for generators/exceptions restored from a backup.
...
Also propagated the sysflag/owner/secclass assignment and validation to all object types and wiped out the duplicating system triggers.
Warning: testing is required!
2013-08-04 15:19:31 +00:00
dimitr
31b4f14055
A quick fix for the bugcheck appeared in test bugs.core_195.
2013-07-15 13:44:47 +00:00
dimitr
8e32f8f5f8
Misc.
2013-07-02 11:27:02 +00:00
dimitr
15e4367810
Small refactoring.
2013-03-17 17:35:53 +00:00
hvlad
50c4099b92
Make tracing VIO works and more easy to use:
...
- log into file instead of console
- rebuild only small set of affected modules when VIO_DEBUG is changed
2013-03-10 14:48:17 +00:00
dimitr
86f4b02ef3
1) Cleanup and renaming in the SCL code.
...
2) Reimplemented the USAGE privilege as a separate one.
3) Added USAGE permission checks for generators/sequences and exceptions.
4) Supported USAGE in GRANT/REVOKE for all object types.
5) Refactored SET GENERATOR and ALTER SEQUENCE as true DDL. blr_set_generator is supported only for backward compatibility, it's not generated by DSQL anymore.
6) Added START WITH clause for [RE]CREATE SEQUENCE and CREATE OR ALTER SEQUENCE.
7) Fixed a number of related errors.
2013-03-07 13:59:03 +00:00
alexpeshkoff
7a9e867ab4
comment
2013-01-08 10:51:02 +00:00
alexpeshkoff
c7454a49a5
Work in progress (committed due to SF upgrade). Checked only on SS.
...
1. Frontported fix for CORE-3935 and CORE-3993.
2. Added debugging support for mutexes and rwlocks in Vulcan style.
Unfortunately after last 'svn up' build asserts in JrdStatement:71.
2012-12-14 17:59:02 +00:00
asfernandes
9727c2685b
Misc.
2012-12-02 01:45:18 +00:00
hvlad
6afc531f86
Make sure relation's flags are loaded from disk before analyze them
2012-11-20 12:23:53 +00:00
asfernandes
6a0a6f2d13
Misc.
2012-11-05 01:14:04 +00:00
alexpeshkoff
4ca6e669fe
1. Fixed CORE-3786: Firebird 2.5.1 Hangs on MacOSX 10.7 (Lion) on DB create after reboot.
...
2. Cleanup: avoid functional style in stuff exception.
2012-11-01 11:40:18 +00:00
dimitr
657511babe
Wiped out the legacy attachment level counters in favor of the already existing runtime statistics. Introduced a record lock counter (more to follow).
2012-11-01 00:16:55 +00:00
dimitr
5847f3860f
Avoid a sweep pass for temporary tables, as the sweeper always sees those tables as empty, but even setting them up as empty has a non-zero cost.
2012-10-31 14:20:24 +00:00
asfernandes
b7d87a6cfa
Misc.
2012-10-16 00:51:42 +00:00
hvlad
513ce5e1ce
Additional patch for improvement CORE-3362 : Cursors should ignore changes made by the same statement.
...
- SubQueryNode (sub-select's) and RseBoolNode (exists(), in\any\all) now knows if it is a part of some outer select (ForNode) or it is a standalone statement.
- VIO\get_undo_data was reworked to better fit undo-log implementation, corresponding changes in VIO_chase_record_version also done
- AV in VIO\get_undo_data was fixed
2012-09-24 14:26:33 +00:00
dimitr
594bae44d2
Fixed CORE-3924: Bugcheck 291 (cannot find record back version) if GTT is modified concurrently using at least one read-committed read-only transaction.
2012-09-12 05:26:38 +00:00
hvlad
b636315561
Improvement CORE-3656 : Support for sweep information in Trace API.
...
Fixed case when sweep cleaned up relation with all records deleted but corresponding event was missed in trace log.
2012-09-04 11:35:16 +00:00
hvlad
9124f5e188
Front ported improvements
...
CORE-3656 : Support for sweep information in Trace API, and
CORE-2668 : Write note into log when automatic sweep is started
Also per-table stats is enabled at transaction level (see thread_db::bumpRelStats)
2012-08-02 08:55:21 +00:00
dimitr
42a4c93b97
1) Removed the no longer actual "parent lock" concept from both the lock and event managers.
...
2) Reworked the lock manager synchronization: added the RAII mutex holders and avoided dirty reads for the shared memory.
3) Encapsulated lck_length into the Lock class + small related refactoring.
4) Front ported recent fixes from the v2.x branches.
5) Misc cleanup.
2012-06-21 15:37:38 +00:00
asfernandes
07557dc297
Misc.
2012-05-26 18:05:56 +00:00
robocop
642d506cd9
Change txn numbers to unsigned and commit after Vlad's review. If the compile-time checks (to ensure enough room for the new txn type) are too ugly, they may be deleted.
2012-05-19 12:04:37 +00:00
robocop
70d5ebdfdf
- Got rid of the fixed array that kept the stream count in the first element and 255 streams after it.
...
- Changed streams from UCHAR to ULONG as Dmitry wanted
- Got rid of some redundant flags to indicate BLR level
- Misc changes (for example, converted to test for DBB_read_only in Database to a function)
- BLR v6 is left commented (waiting for a better solution to appear, if any exists)
- This change is not going to raise the engine limits until a solution is found to express more than 255 streams in BLR.
2012-04-12 09:02:13 +00:00
hvlad
24d152138d
Refactoring of runtime statistics and related classes:
...
- moved relation's counters into RuntimeStatistics.h
- renamed RuntimeStatistics::bumpValue to RuntimeStatistics::bumpRelValue and thread_db::bumpStats to thread_db::bumpRelStats to make things more clear
- removed not obvious cast in VIO_bump_count
- removed dependency of RuntimeStatistics.h from ntrace.h
2012-04-11 09:38:19 +00:00
alexpeshkoff
4e795f4047
Include platform definitions file common.h in firebird.h. Use OS-specific ThreadId instead artificial FB_THREAD_ID, avoiding unneeded casts in the code. This should fix BSD and MAC ports.
2012-03-01 08:55:43 +00:00
robocop
21cd78a0fd
Cleanup: unused functions, parameters & variables.
2012-01-24 03:37:43 +00:00
dimitr
f0112290c6
Disallowed deletions from RDB$DATABASE. This fixes the only problematic part of CORE-3735: Unprivileged user can delete from RDB$DATABASE, RDB$COLLATIONS, RDB$CHARACTER_SETS. The implemented solution could be used to protect other system tables from direct modifications.
2012-01-20 10:23:31 +00:00
asfernandes
c44c548f8c
Corrections and cleanup.
2012-01-09 00:08:33 +00:00
dimitr
e956e2e6c0
1) Restricted ALTER/DROP permissions to the object owners only. Now this applies to domains, charsets, collations, generators and exceptions. This fixes CORE-304: Metadata security hole - any user can alter/drop generators and exceptions.
...
2) Opened the gates to implement the standard USAGE privilege (CORE-2884). SQL support and validation logic are still to be developed.
3) Added the grant option to the owner permissions for packages, procedures and functions.
4) Misc cleanup and refactoring.
2012-01-08 14:05:29 +00:00
dimitr
aa4e36e0d6
Correction.
2011-12-01 09:37:25 +00:00
dimitr
e3fb73cdca
Attempted to fix CORE-3490: Concurrency problem when using named cursors. The solution may look sub-optimal but it doesn't seem to affect the performance.
2011-12-01 09:24:59 +00:00
dimitr
d542f5ac88
Fixed the races leading to a crash during dbb finalization. Thanks to Adriano for assistance.
2011-11-01 04:02:59 +00:00
asfernandes
906ea4729e
Misc.
2011-10-27 01:04:14 +00:00
dimitr
444d5302e1
Refactored the tdbb initialization, especially in AST and BG threads.
...
Replaced dbb_use_count with a list of active requests (to be utilized later).
Misc cleanup.
2011-10-18 18:02:57 +00:00
asfernandes
eb5689ef0e
Misc.
2011-10-07 02:33:02 +00:00
hvlad
ca1a67992c
Return explicit sign of active transaction from VIO_get_current() - it makes logic more clear in check_duplicates() and allows to clear transient lock error from status vector and not propagate up to the callers.
...
Cleanup : removed my old disabled code (checking for old and new values) and restored original code in check_duplicates().
2011-10-03 11:27:46 +00:00
hvlad
dc4abc05e4
Protect system attachments list with dedicated sync object. It prevents deadlocks with common sync object on process shutdown.
2011-10-02 10:19:06 +00:00
hvlad
fb36296224
Port forward fix for bug CORE-3610 : Can insert DUPLICATE keys in UNIQUE index
2011-09-29 12:34:07 +00:00
dimitr
b100b395eb
Being a little bit more paranoid doesn't hurt.
2011-09-22 18:15:21 +00:00
dimitr
031339406c
Methinks it's more correct from the scoping POV.
2011-09-21 17:49:49 +00:00
dimitr
809fba5ec9
Reworked the garbage collector thread:
...
- a single try-catch block looks better
- cross-helping between GC and CW/CR threads is agreed to be a bad idea
- replaced ATT_disable_gc_notify with a check inside notify_garbage_collector()
- made the logic more straightforward
Careful review would be appreciated.
2011-09-21 07:59:15 +00:00
hvlad
652200448a
Port forward fix for bug CORE-3557 : AV in engine when preparing query against dropping table
2011-07-16 10:41:27 +00:00
asfernandes
47489fa71b
Misc.
2011-06-26 18:48:00 +00:00
robocop
1c250c02eb
Misc.
2011-06-25 03:40:18 +00:00
hvlad
f12c243e1c
Improvement CORE-3537 : There is no need to undo changes made in GTT created with ON COMMIT DELETE ROWS option when transaction is rolled back.
2011-06-24 12:54:46 +00:00
hvlad
f2aaa12ce5
Improvement CORE-3536 : Garbage collection in GTT is unnecessary delayed by active transactions in another attachments
2011-06-24 12:49:08 +00:00
hvlad
7c68e15a0f
Delete deferred work when savepoint is cleaned up after failure even if no records was modified. Else DFW task will be repeated at commit making system tables logically inconsistent.
2011-06-24 12:03:32 +00:00
robocop
5ad96271e6
Misc.
2011-06-24 06:34:16 +00:00
hvlad
81cec16f83
Fixed leak of SysAttachment instance. Noted by Claudio.
2011-06-13 11:09:19 +00:00
asfernandes
add3d8ddef
Misc.
2011-06-08 01:55:13 +00:00
hvlad
945a1bdbf4
Fixed bug CORE-3515 : Index corruption. Validation put "missing entries" message into firebird.log
2011-06-05 08:15:33 +00:00
dimitr
df893a708d
Better (I hope) solution for CORE-3058: New generators are created with wrong value when more than 32K generators was previously created. It also resolves the [temporary] inability to restore generator values.
2011-06-01 16:58:55 +00:00
asfernandes
2c9dd7e1ff
Misc.
2011-05-28 02:05:45 +00:00
hvlad
99c9c8e2db
Enabled background threads (garbage collector and cache writer).
...
Introduced special kind of system attachments used in this threads and show its activity in monitoring.
2011-05-27 07:57:16 +00:00
asfernandes
3ce54d2eda
Misc.
2011-05-11 01:18:28 +00:00
hvlad
fef1f49c94
Shared page cache implementation
2011-05-09 10:15:19 +00:00
dimitr
4b3033c013
Better locality of allocations. This also fixes a tiny memory leak in SuperServer.
2011-04-05 05:36:05 +00:00
robocop
82403a5621
Misc, style, etc.
2011-04-02 04:45:26 +00:00
asfernandes
e4658d76c3
Misc
2011-03-20 16:24:46 +00:00
hvlad
58a55cc813
Improvement CORE-3399 : Allow write operations to temporary tables in read only transactions
2011-03-19 13:10:51 +00:00
dimitr
d68affa114
More complex fix for CORE-3374: Server may crash or corrupt data if SELECT WITH LOCK is issued against records not in the latest format. It involves upgrading the record to the current format, like UPDATE does.
2011-03-05 17:33:01 +00:00
asfernandes
72b7e83082
Misc
2011-02-26 19:05:09 +00:00
asfernandes
87ec1ad31c
The (bad) definition of fb_assert cause this code to not work in clang. It's a surprive if it works in MSVC.
2011-02-25 15:28:10 +00:00
hvlad
07e245ab1b
Implement improvement CORE-3362 : Cursors should ignore changes made by the same statement
2011-02-25 12:26:57 +00:00
dimitr
7a405a4eea
Fixed CORE-3058: New generators are created with wrong value when more than 32K generators was previously created.
2011-02-17 06:05:16 +00:00
alexpeshkoff
f89ac854ef
Getting ready to use 32 (not 31) page and transaction numbers
2011-01-21 17:55:04 +00:00
asfernandes
5c7e5ce634
Misc
2010-10-22 01:24:31 +00:00
hvlad
f223086fdd
Corrections for new build system. Work in progress...
2010-10-18 17:23:36 +00:00
dimitr
af23581244
Front ported my recent changes. I'm unable to verify the compilation now, sorry.
2010-10-14 05:27:08 +00:00
alexpeshkoff
e05c1a825f
Moved some more .h-files to common from jrd as Vlad suggested
2010-10-13 10:39:52 +00:00
alexpeshkoff
83d6a2f88c
Restored a lot of changes, rolled back by me unintentionally
2010-10-12 11:36:51 +00:00
alexpeshkoff
2dc25a44bb
OSRI and new posix build
2010-10-12 08:02:57 +00:00
dimitr
34dd100554
Fixed CORE-3137: Partial rollback is possible for a selectable procedure modifying data.
2010-09-18 06:52:46 +00:00
asfernandes
96cd57ba83
Better error reporting and adjustments
2010-07-29 00:36:31 +00:00
asfernandes
36f092fe0e
Improvement CORE-3085 - Add clause ALTER DOMAIN <name> [NOT] NULL
2010-07-28 16:14:20 +00:00
dimitr
60763936ac
Fixed the second part of the problem described in CORE-2933: Very slow execution of a script that creates a lot of metadata.
2010-05-17 09:58:45 +00:00
asfernandes
d6e4b53865
Misc
2010-05-08 02:25:58 +00:00
skidder
c84d62d82e
Fix CORE-2988. Concurrent transaction number not reported if lock timeout occurs
2010-05-04 17:29:44 +00:00
asfernandes
4763533669
Refactor jrd_req into JrdStatement (shared part) and jrd_req (execution of a statement).
...
Fixed some bugs found in the way.
2010-04-18 22:19:11 +00:00
alexpeshkoff
88b8a40a04
1. Cleaned up most of SUPERSERVER usage.
...
SUPERSERVER has 4 different meanings in our code:
- this is milticlient server (replaced with Config::getMultiClientServer())
- use shared data cache (replaced with Config::getSharedCache())
- use shared metadata cache (replaced with macro SHARED_METADATA_CACHE)
- database are NOT shared between processes (replaced with getSharedDatabase())
2. Use fb_smp_server as both superclassic and classic binary on posix
(xinetd autodecection is done).
3. Small posix build cleanup.
2010-04-15 14:40:27 +00:00
asfernandes
e1ba33a2ae
Fix problem of the system transaction storing pointer to random or already deallocated attachment. Full build (or kind of) required.
2010-04-02 21:48:15 +00:00
hvlad
521f7a5cbc
Fixed bug CORE-2879 : Sweep could raise error : page 0 is of wrong type (expected 6, found 1)
2010-02-22 11:09:59 +00:00
robocop
3b5a0fdae5
The way AccessItem instances were ordered was a hack that sooner or later would fail when adding more object types (that now have ACLs but we still don't enfornce them). Further, comparing numbers is faster than comparing strings.
2010-02-13 08:31:16 +00:00
dimitr
4c4a77a4fe
Refactored SQZ into a class. Fixed all the SCHAR* pointers to be UCHAR* + appropriate cleanup.
...
Make the interface and internals friendly to any byte sequence, even the one longer than 32K.
2010-02-12 12:53:06 +00:00
robocop
656dcdb1a5
Got rid of some casts. Please double check it as it's critical for data. The only trick seems to be taking the value of a byte as a signed char, so it can become negative.
2010-02-12 11:29:51 +00:00
robocop
753f914fb1
CORE-2858: Possible memory trashing when raising some exceptions to signal failed security checks. Thanks Dmitry.
2010-02-12 11:22:22 +00:00
robocop
1d87d749f0
Misc.
2010-01-26 08:20:27 +00:00
asfernandes
2faab16698
Misc
2010-01-21 02:51:32 +00:00
hvlad
258f322f67
Avoid possible deadlock in VIO.
2010-01-19 18:41:44 +00:00
hvlad
abdfa188d5
Preserve record's flags when gathering record from fragments. Its used in prepare_update() to check if primary record was backed out.
2010-01-19 18:38:52 +00:00
asfernandes
e8cc188227
Store the view context type when it doesn't came from backup and simplify the code. Still problems altering procedures used in views.
2009-12-28 15:31:09 +00:00
asfernandes
d8c99a9be2
1) Make procedures and functions member of the same hierarchy (Routine class)
...
2) Some completion for external functions
2009-12-27 22:05:22 +00:00
