| Commit message (Collapse) | Author | Age | Files | Lines |
|\ |
|
| |\ |
|
| | |\
| | | |
| | | | |
WT-2503 Rename "wait" to avoid shadowing system call.
|
| | | | |
|
| | |\ \
| | | | |
| | | | | |
WT-2506 Reorder to fix uninitialized variable warning.
|
| | | |/ |
|
| | |\ \
| | | |/
| | |/| |
WT-2501 LSM drop can violate access assumptions.
|
| | | | |
|
| | | |\
| | | |/
| | |/| |
|
| | |\ \
| | | | |
| | | | | |
WT-2499 Fix a race in LSM between setting and checking WT_LSM_TREE_ACTIVE
|
| | | | | |
|
| | | | |
| | | | |
| | | | |
| | | | | |
We should only set active when not opening an exclusive handle.
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Tree close relies on any state change being visible immediately. Found
by inspection, but could cause the symptom seen in this ticket.
|
| | |\ \ \
| | | | | |
| | | | | | |
WT-2502 Free local allocation.
|
| | | | | | |
|
| | |/ / / |
|
| | |\ \ \
| | | |/ /
| | |/| | |
WT-2426 Lock all handles for checkpoints in one pass.
|
| | | |\ \
| | | |/ /
| | |/| | |
|
| | | |\ \
| | | | | |
| | | | | | |
WT-2496 Fix a race between checkpoint and starting a hot backup.
|
| | | |/ /
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
In particular, a checkpoint could mark and lock an old checkpoint for
deletion before a hotbackup starts. The hotbackup could then write
metadata referencing that old backup which the checkpoint goes on to
delete.
The fix is crude: don't allow a hotbackup to start while a checkpoint
is in progress.
|
| | | |\ \ |
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
In particular, gather all locks (including exclusive locks on old
checkpoints) under the schema / handle list lock so we never reenter
those locks once we have a handle exclusive.
This extends the duration that we hold exclusive locks on checkpoints
to include the write leaves phase. If that turns out to be a problem,
we could further split out write leaves into a separate pass (that
releases all handles at the end).
|
| | | |_|/
| | |/| |
| | | | |
| | | | | |
Add a test case demonstrating the problem.
|
| | |\ \ \
| | | | | |
| | | | | | |
WT-2497 Create second backup copy.
|
| | | | | | |
|
| | |\ \ \ \
| | | |/ / /
| | |/| | | |
WT-2498 Fix a bug in LSM where table drop could hang.
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Always wait for internal operations, optionally wait for application
operations.
|
| | |/ / /
| | | | |
| | | | |
| | | | | |
If there are open cursors.
|
| | |\ \ \
| | | | | |
| | | | | | |
WT-2490: search_near() returns wrong key for column-store
|
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
This reverts commit 6bf871c9cc0379a15bf1aa550df6a1ea2dfd4521.
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
In variable-length column-store, updates/deletes are stored in the
insert list. (In short, there needs to be a place to put references for
individual records because we might update/delete a single entry in a
large duplicate set, and the insert list has to exist anyway, so we
used it.) However, if we find a match in the insert list, but it's not
visible to the current transaction for some reason, we need to know if
there's an on-page record with the same record number with a previous
version of the data we can use instead.
We were using an illegal cursor slot value for this purpose, but that
messes up search-near, which needs a valid slot should the entry on
the insert list not be visible.
Add a new flag to the btree cursor, WT_CBT_VAR_ONPAGE_MATCH, that's set
if we find a matching on-page variable length column-store record.
Clarify the column-search routines, specifically set the cursor's
slot and record number for each code path.
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Enhance the variable-length column-store search/search-near tests
to catch WT-2490.
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Lint: Last value assigned to variable 'firstcg_cur' not used in one code
path.
|
| | | |\ \ \ |
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Update string list.
|
| | |\ \ \ \ \
| | | | | | | |
| | | | | | | | |
WT-2495 Make malloc calls use tcmalloc when configured.
|
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
Otherwise we can free memoy with tcmalloc that was allocated
with the system allocator.
|
| | |\ \ \ \ \ \
| | | |_|_|_|_|/
| | |/| | | | | |
WT-2414 Avoid main table reads and extractions for the first index
|
| | | | |_|/ /
| | | |/| | |
| | | | | | |
| | | | | | | |
the index key for the first index available. We don't need to extract it again.
|
| | |\ \ \ \ \
| | | |_|/ / /
| | |/| | | | |
WT-2457 Wait for LSM work units to drain when getting exclusive access.
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Rename LSM field "excl_session" to match WT_DATA_HANDLE, make new test
URI match test name.
|
| | | | | | | |
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Otherwise we can return EBUSY unexpectedly. Fixed several outstanding
bugs along the way.
|
| | |\ \ \ \ \
| | | |_|_|_|/
| | |/| | | | |
WT-2492 Change printf format specifier to be Windows compatible.
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Since there isn't a specifier that works on all platforms.
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Windows doesn't like %td, use the maximum unsigned integer format
instead, PRIuMAX.
Get rid of the temporary variable I introduced in cc0d34b, there's
no need for one.
|
| | | | |/ /
| | | |/| |
| | | | | |
| | | | | | |
Windows doesn't like "td" for ptrdiff_t
|
| | |\ \ \ \
| | | | | | |
| | | | | | | |
WT-2481 Set memory_page_max after we read chunk_max.
|
| | | | | | | |
|
| | |\ \ \ \ \
| | | | | | | |
| | | | | | | | |
WT-2485 Adjust lsm_tree->merge_min when we read the metadata.
|