Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
100 check-ins occurring around 15abbc34168f7a5b.
|
2010-06-03
| ||
| 19:10 | Fix another problem in test_vfs.c. (check-in: df7d59899c user: dan tags: trunk) | |
| 18:20 | Have test_vfs.c simulate IO errors in xShmLock. (check-in: fcbf7cf189 user: dan tags: trunk) | |
| 18:02 | Performance fix for winShmClose(). (check-in: ed7774de04 user: drh tags: trunk) | |
| 16:58 | Add extra tests for removing elements from wal-index hash tables as part of a rollback. (check-in: af3e598ad9 user: dan tags: trunk) | |
| 12:35 | Delay opening the sub-journal until SQLite actually needs to write data to it. (check-in: c43deb33ae user: dan tags: trunk) | |
| 12:09 | Remove global variables when compiled with SQLITE_OMIT_WSD (check-in: dd10a547f1 user: drh tags: trunk) | |
| 09:25 | Enhancements to test_vfs.c and walfault.test. (check-in: ac0de2f39e user: dan tags: trunk) | |
| 09:17 | If an error (OOM or SQLITE_FULL error) occurs while executing an SQL statement and a statement-transaction is automatically rolled back as a result, if a second error occurs during the statement rollback do a full transaction rollback instead. Otherwise the client can be left with an inconsistent cache. This can affect both WAL and rollback modes. (check-in: eb80ddc665 user: dan tags: trunk) | |
| 09:01 | If a malloc fails while allocating a savepoint object at the pager level, do not try to roll that savepoint back later on. (check-in: 91cb08ffb6 user: dan tags: trunk) | |
|
2010-06-02
| ||
| 20:38 | When a lock fails due to an I/O error (not an SQLITE_BUSY) then the checkpoint should fail. (check-in: c6eb0a2727 user: drh tags: trunk) | |
| 18:59 | Fix a problem with rolling back to a savepoint opened before the writer decided to wrap the log file. (check-in: 6b4aed6aae user: dan tags: trunk) | |
| 17:15 | If an error occurs while writing frames to the log to commit a transaction, do not mark the pages as clean in the cache. Otherwise, the subsequent rollback does not roll them back (leaving the client to continue with a cache that makes it appear that the transaction was committed - inconsistent with the database on disk). (check-in: cb571c1b71 user: dan tags: trunk) | |
| 16:40 | Added implementation of vfslog_time() for Windows testing. (check-in: b3109f85bd user: shaneh tags: trunk) | |
| 16:39 | Updated wal2.test for windows testing. (check-in: c3606f3985 user: shaneh tags: trunk) | |
| 15:59 | Make sure the directory used by os_unix.c for shared-memory files is always the same, regardless of individual client permissions and environment variable settings. (check-in: f97bd520f0 user: drh tags: trunk) | |
| 15:43 | When walTryBeginRead() encounters an I/O error trying to set a lock (as opposed to SQLITE_BUSY) be sure to propagate that error back up the call stack. (check-in: aa2c2b67a7 user: drh tags: trunk) | |
| 14:45 | Changes to make WAL more robust against SHM locking failures and OOM errors. (check-in: ebf4041383 user: drh tags: trunk) | |
| 14:43 | In the rowhash.test, make sure global variables are cleared prior to use. (check-in: 28efe0a404 user: drh tags: trunk) | |
| 12:58 | Undo one of the "const" markers from the previous check-in because the value needs to be changable, even if it is not actually ever changed in the current implementation. (check-in: c24ac3563f user: drh tags: trunk) | |
| 05:53 | Add some 'const' markers to static data that is really constant. (check-in: e7073e23b8 user: dan tags: trunk) | |
|
2010-06-01
| ||
| 21:02 | The shared-memory used by WAL on linux now really is shared memory in /dev/shm. On other unix flavors, the file is in a temporary directory rather than in the same directory as the database. (check-in: fc18c4aadb user: drh tags: trunk) | |
| 19:15 | Changes to the way faults are injected into xShmXXX VFS calls. (check-in: 716d99f392 user: dan tags: trunk) | |
| 17:46 | Change the OOM and IO error test cases in walfault.test so that each test case runs both types of error simulation. (check-in: b627e15368 user: dan tags: trunk) | |
| 15:44 | Delay the decision to restart the log file until data is actually ready to be written to the log file (instead of at the start of a write transaction). (check-in: b1abfaaf53 user: dan tags: trunk) | |
| 15:24 | The incremental checkpoint feature is not perfect yet, but it is working well enough to merge it into the trunk. (check-in: 1d3e569e59 user: drh tags: trunk) | |
| 14:30 | Merge with [15abbc3416]. (Closed-Leaf check-in: f4b9003a2d user: dan tags: wal-incr-ckpt) | |
| 14:12 | If an attempt to get a read-lock on the WAL fails with SQLITE_BUSY_RECOVER, call the busy-handler at the btree level. (check-in: ce64496509 user: dan tags: wal-incr-ckpt) | |
| 13:17 | Fix a bug introduced by the previous check-in but only seen if SQLITE_ENABLE_EXPENSIVE_ASSERTS is turned on. (check-in: 9c9ec8994b user: drh tags: wal-incr-ckpt) | |
| 12:58 | Fix an off-by-one boundary-value issue in walCleanupHash(). (check-in: f039552e63 user: drh tags: wal-incr-ckpt) | |
| 11:08 | Do not hold the shared-memory mutex in os_unix.c if returning NULL. (check-in: 9622dd468c user: dan tags: wal-incr-ckpt) | |
| 10:44 | If the checkpoint fails to obtain an exclusive lock on one of the read-lock bytes, do not consider this an error. (check-in: 9e95e35728 user: dan tags: wal-incr-ckpt) | |
| 07:51 | Fixes to the test cases in wal2.test. (check-in: cd5fbcbce8 user: dan tags: wal-incr-ckpt) | |
| 01:08 | Comment edits and cleanup in wal.c. No functional code changes. (check-in: e8e666ab82 user: drh tags: wal-incr-ckpt) | |
| 00:28 | Attempt to get the filectrl.test script running. (check-in: e46a8f2b75 user: drh tags: wal-incr-ckpt) | |
| 00:03 | Fix os_unix.c so that it will compile and build on a Mac. (check-in: bc707c83e5 user: drh tags: wal-incr-ckpt) | |
|
2010-05-31
| ||
| 20:28 | Fix issues with locking_mode=EXCLUSIVE in WAL. (check-in: 8deba0cebd user: drh tags: wal-incr-ckpt) | |
| 18:24 | Add an "isInit" field in the wal-index header that must be non-zero for a valid header. Use this to detect an uninitialized wal-index. (check-in: a16fde1901 user: drh tags: wal-incr-ckpt) | |
| 16:56 | Fix an inconsistent #ifdef in wal.c. Fix os_unix.c so that it does not allow moving an SHM lock directly exclusive to shared without going through unlocked. (check-in: 552658da28 user: drh tags: wal-incr-ckpt) | |
| 16:41 | Avoid dropping the checkpoint lock after a recovery run as a precursor to a checkpoint operation. (check-in: cc25cfa046 user: dan tags: wal-incr-ckpt) | |
| 16:17 | Zero the checkpoint header as the last step of successful WAL recovery. Avoid an unnecessary lock/unlock in WalBeginReadTransaction. (check-in: db3509c55d user: dan tags: wal-incr-ckpt) | |
| 16:10 | Get the new xShmLock interface design working on os_win.c. (check-in: 149a7082e2 user: drh tags: wal-incr-ckpt) | |
| 14:39 | Make sure WAL alway requests enough shared-memory space. (check-in: 138f128317 user: drh tags: wal-incr-ckpt) | |
| 14:28 | Remove the tests that enable tracing based on the existance of the "vdbe_*" files when in SQLITE_DEBUG mode. (check-in: cdbb4e7ca7 user: drh tags: wal-incr-ckpt) | |
| 14:18 | Update another test case in wal.test. (check-in: 99fde69e61 user: dan tags: wal-incr-ckpt) | |
| 13:11 | Change WAL test cases to account for the improved concurrency in the new checkpoint logic. (check-in: 2d6f248505 user: drh tags: wal-incr-ckpt) | |
| 12:30 | Do not fail a checkpoint just because active readers prevent backfill. (check-in: 9aa4243e0c user: drh tags: wal-incr-ckpt) | |
| 12:15 | Add WALTRACE() macros to help with debugging and analysis. (check-in: 765c33181a user: drh tags: wal-incr-ckpt) | |
| 12:02 | Another minor fix to checkpoint on the WAL branch. (check-in: 853f1e3f77 user: dan tags: wal-incr-ckpt) | |
| 11:39 | Fix a bug in checkpoint introduced by [181ceb32ea]. (check-in: b499dbc88a user: dan tags: wal-incr-ckpt) | |
| 11:16 | Fix some asserts and other things in the new WAL branch. (check-in: 181ceb32ea user: dan tags: wal-incr-ckpt) | |
| 06:38 | Changes to the way one of the WAL/OOM tests works. (check-in: 15abbc3416 user: dan tags: trunk) | |
| 01:41 | WAL runs but quickly deadlocks. (check-in: ace58acbf1 user: drh tags: wal-incr-ckpt) | |
|
2010-05-30
| ||
| 19:55 | Initial code for incremental checkpoint in WAL mode. This check-in compiles on unix and runs as long as you do not engage WAL mode. WAL mode crashes and burns. Consider this check-in a baseline implementation for getting the new capability up and running. (check-in: ef3ba7a17f user: drh tags: wal-incr-ckpt) | |
|
2010-05-29
| ||
| 08:40 | Add tests to fkey2.test to check that ON CONFLICT clauses do not affect SQLite's behaviour when an FK constraint is violated. (check-in: e9e5b10019 user: dan tags: trunk) | |
| 06:18 | Minor changes to checksum related test cases. (check-in: 60c22bde52 user: dan tags: trunk) | |
|
2010-05-28
| ||
| 11:16 | Add a test to walcksum.test to stress the checksum calculation. (check-in: 778d0c1768 user: dan tags: trunk) | |
|
2010-05-26
| ||
| 17:31 | Make sure the wal-index mapping is always large enough to cover the entire active area of the wal-index. (check-in: 42705babba user: drh tags: trunk) | |
| 15:06 | Change the semantics of xShmGet() such that it will never increase the size of shared memory. xShmSize() must be used to grow the size of shared memory. A shared memory segment size cannot be shrunk (except by dropping it). (check-in: 72de007312 user: drh tags: trunk) | |
|
2010-05-25
| ||
| 15:53 | Updated header comments in wal.c. No functional code changes. (check-in: 687632a6b3 user: drh tags: trunk) | |
| 15:23 | Add a busy-handler to a test case in walthread.test to prevent errors. (check-in: d3d348aa97 user: dan tags: trunk) | |
| 13:49 | Modify walcrash2.test so that it works with DEFAULT_AUTOVACUUM=1. (check-in: 77438882dd user: dan tags: trunk) | |
| 13:40 | Update header comments in wal.c to correctly describe the WAL file format. Update the locking region offsets in os_unix.c and os_win.c and add assert() statement to verify that the locking region offsets are correct. (check-in: 40030c0739 user: drh tags: trunk) | |
| 10:50 | If a writer exits unexpectedly in the middle of a transaction, have the following writer remove any wal-index hash-table entries left by the interrupted transaction. (check-in: ed77556adc user: dan tags: trunk) | |
| 02:24 | Remove unreachable code associated with WAL from the pager. (check-in: 54c1718e6d user: drh tags: trunk) | |
|
2010-05-24
| ||
| 20:27 | OOM errors during an auto-checkpoint are benign. (check-in: 3d252ce5d0 user: drh tags: trunk) | |
| 20:24 | Remove the walmode.test case from the inmemory_journal test of permutations.test. (check-in: 27ba9c21fb user: drh tags: trunk) | |
| 17:00 | Fix the wal2.test script so that it works in auto_vacuum mode. (check-in: 6a818afb93 user: drh tags: trunk) | |
| 13:57 | Change the checksum used in WAL files so that each frames checksum depends on the content of the WAL header and all frame headers and content up to and including the frame to which the checksum is attached. (check-in: 8a53f12c83 user: dan tags: trunk) | |
| 13:28 | Make sure a WAL frame of all zeros is detected as an invalid frame. (check-in: 02d99ad4b5 user: drh tags: trunk) | |
| 12:34 | Fix up test_osinst.c to work with SQLITE_OMIT_VIRTUALTABLE. (check-in: 51fd38152b user: drh tags: trunk) | |
| 10:39 | Change the WAL file format to support two kinds of checksums - one that is fast to calculate on little-endian architectures and another that is fast on big-endian architectures. A flag in the wal-header indicates which the file uses. (check-in: 65ba804dd1 user: dan tags: trunk) | |
|
2010-05-22
| ||
| 08:22 | Add a couple of missing methods to test_osinst.c.. (check-in: 5c9e9c06ae user: dan tags: trunk) | |
| 00:55 | Add several EXPENSIVE_ASSERT code blocks to validate the wal-index hash table. Fix the bugs that these code blocks fine. Rename walClearHash() to walCleanupHash() and simplify its interface. (check-in: 7aade899e5 user: drh tags: trunk) | |
|
2010-05-21
| ||
| 19:15 | Fix another bug in walClearHash(). (check-in: 40f80ffe70 user: dan tags: trunk) | |
| 18:24 | Comment out the SQLITE_FCNTL_SIZE_HINT from os_unix.c since it does not seem to provide any performance gain there. (check-in: 7d01309da6 user: drh tags: trunk) | |
| 16:23 | Correct an off-by-one bug in the previous commit. (check-in: 75a1130d86 user: dan tags: trunk) | |
| 15:31 | Remove entries from wal-index hash tables when a rollback or savepoint rollback occurs. (check-in: 36795c2b23 user: dan tags: trunk) | |
| 13:16 | Fix the wal-index header read routine so that it correctly detects a zero header as being malformed and in need of a wal-index rebuild. (check-in: 1a4eb3a3ef user: drh tags: trunk) | |
|
2010-05-20
| ||
| 23:51 | Add a new xShmBarrier method to the VFS - a shared-memory fence operation. Implement the same in both unix and win32. Use it to make the WAL subsystem more robust. (check-in: 1bd011c9fe user: drh tags: trunk) | |
| 21:21 | Make use of the extra information in the WAL header and frame header to enhance robustness. (check-in: 9580ecb7e3 user: drh tags: trunk) | |
| 16:45 | Convert the wal-header and frame-header to 24 bytes. Extra information in both headers is designed to enhance robustness after crashes, though the extra information is currently unused. This is a snapshot of a work in progress. (check-in: 669706431f user: drh tags: trunk) | |
|
2010-05-19
| ||
| 23:41 | Merge WIN32 WAL support into trunk. Still some issues with locking to resolve. (check-in: 43377663fc user: shaneh tags: trunk) | |
| 22:09 | Cherry-pick the SQLITE_FCNTL_SIZE_HINT patch (check-in [2b7e3b4a30d6a7c4a8] and bump the version number to 3.6.23.2. (check-in: 776679af58 user: drh tags: branch-3.6.23) | |
| 20:30 | Disable the corruptE test which SQLITE_OVERSIZE_CELL_CHECK is enabled. This is a cherrypick merge from check-in [1dc4f9fe9b83467ba]. (check-in: a40d080ca3 user: drh tags: branch-3.6.23) | |
| 19:39 | Add the SQLITE_FCNTL_SIZE_HINT to the os_win.c VFS. (check-in: 9894118d20 user: drh tags: trunk) | |
| 19:26 | Add the SQLITE_FCNTL_SIZE_HINT operator to sqlite3_file_control() and use it to give the VFS hints about the ultimate size of a database file when the file is growing. (check-in: 2b7e3b4a30 user: drh tags: trunk) | |
| 19:09 | Fix a bug in the new checkpoint computation. Also update the checkpoint algorithm in the test scripts to align with the new implementation. (check-in: 8b6056f2ee user: drh tags: trunk) | |
| 18:08 | Revise the checksumming algorithm in wal.c. More variable refactoring. (check-in: 542b90eba6 user: drh tags: trunk) | |
| 17:49 | Refactoring some variable names in wal.c. (check-in: 1d201ff51f user: drh tags: trunk) | |
| 01:53 | Add a large comment to wal.c describing the WAL and wal-index file formats. (check-in: a71a22b52f user: drh tags: trunk) | |
|
2010-05-18
| ||
| 23:29 | Update the wal-index hash format so that hash-table space is reused following a rollback, thus preventing hash table overflows. Add assert()s to verify that hash tables do not overfill. Further refactoring of the wal-index code. (check-in: ada9a8c7b6 user: drh tags: trunk) | |
| 18:01 | Refactoring of the WalIterator implementation. (check-in: b5b60fdcc5 user: drh tags: trunk) | |
| 13:27 | Mark the shared-memory in the WAL implementation as volatile. (check-in: 0a6787908e user: drh tags: trunk) | |
| 12:56 | Comment clarifications in wal.c. (check-in: a029be1017 user: drh tags: trunk) | |
| 12:37 | Keep additional state information in the pager in an effort to reduce the number of SQLITE_FCNTL_SIZE_HINTs. (Note: This change was found to make no performance difference and so has been moved onto a side branch. Use the parent check-in instead of this one.) (Closed-Leaf check-in: 97d88a86e4 user: drh tags: branch-3.6.1-exp) | |
|
2010-05-17
| ||
| 20:16 | Remove an unreachable test from wal.c. (check-in: 7162c45673 user: drh tags: trunk) | |
| 15:52 | An improvement to the SQLITE_FCNTL_SIZE_HINT change that invokes the hint less often and only when really needed. (check-in: a1d20ceb9c user: drh tags: branch-3.6.1) | |
| 15:33 | Invoke the SQLITE_FCNTL_SIZE_HINT opcode on the sqlite3_file_control() interface for database files before extending the size of the file. The VFS can use this hint to preallocate space. (check-in: 9a08371171 user: drh tags: branch-3.6.1) | |
|
2010-05-15
| ||
| 10:24 | Add a note to the documentation of sqlite3_wal_hook() to say not to return SQLITE_ROW or SQLITE_DONE. Or any other value that is not a valid SQLite error code. (check-in: 2f0b6cbf7b user: dan tags: trunk) | |
| 01:09 | Silently ignore errors trying to set the default journal_mode while doing an ATTACH. (check-in: b9af4ad2ef user: drh tags: trunk) | |