ddf24ad848
Validation corrections.
10 years ago
8af2bc551a
Fix #236 via workaround for MySQL issue http://bugs.mysql.com/bug.php?id=75615
10 years ago
34e58dfb2b
Minor cleanup.
10 years ago
c24e6d56bb
Fix #236 use a synchronous Executor to execute setNetworkConnection() to avoid race conditions.
10 years ago
2eefb6d4cd
Checkpoint check-in, this is not final code at all...
10 years ago
b544c21cf3
Add validationTimeout.
10 years ago
898206f81c
Fix #231 Make the copyState() method public to facilitate wrappers who want to copy/modify the configuration for use in multiple pool instances.
10 years ago
28438642d8
Make initializationFailFast fail faster.
10 years ago
1a99e4ca2c
We have the driver as member, so let's use it.
10 years ago
8d58c28077
Remove use of method unsupported by Java 6
10 years ago
5ea5688db3
Start the housekeeping thread first run after the default interval (30 seconds) rather than hardcoded 10 seconds.
10 years ago
684d0683e9
Idle fill is now run by the housekeeping thread, rather than as a side-effect of addBagItem(), so for this test we need to set a short housekeeping period.
10 years ago
3ccddff117
Reset housekeeping period after test.
10 years ago
dc5771e679
Suppress IDE deprecation warnings.
10 years ago
94be387dcc
Clarify boolean order of operations in getConnection() conditional. Run fillPool() on the caller thread.
10 years ago
8aa2034709
Fixes #222 separate responsibility of maintaining idle connections from addBagItem(). addBagItem() will now add exactly one connection per-invocation (with retry). Create a separate fillPool() method to handle idle connection maintenance.
10 years ago
f6b0ee5ab9
#215 don't return evicted connections ever
10 years ago
fafcd38a56
Fix #220 continue to override isClosed() ... we had removed it but it should reflect HikariCP's understanding of the closed state rather than delegating to the driver.
10 years ago
6017c8019b
Fix #221 handle getResultSet()
10 years ago
5de373148c
Don't throw exceptions that might disrupt executor threads, but do log. Plus, don't trust nanoTime() implementations on all platforms not to go backwards due to NTP adjustments.
10 years ago
937923666e
Don't reset isNetworkTimeoutSupported here, but do log a warning on exception
10 years ago
c3043d8f9d
Re-use the assassin executor.
10 years ago
96a71da526
If tryAcquireSharedNanos() returns false, we timed out, just return immediately.
10 years ago
e5d4cf4f11
Change longValue() call to get()
10 years ago
3f64f5c432
JavaDoc fixes.
10 years ago
3fbcbf87b5
When resetting the network timeout to its original value we can avoid an additional call to
...
getNetworkTimeout() by adding a specialized version of the method.
10 years ago
f97d66e554
Move GlobalPoolLock and PoolUtilities into the pool package.
10 years ago
f4818f535e
Expose suspend/resumePool on the HikariDataSource (not just via JMX)
10 years ago
a786a6be32
Loosen visibility requirements on isNetworkTimeoutSupported and isQueryTimeoutSupported, at the minor cost of initial re-execution by different threads.
10 years ago
ea1c2c590a
Slightly kinder, gentler shutdown.
10 years ago
6b63583b0e
Minor cleanup.
10 years ago
b9edd7de18
Minor cleanup.
10 years ago
3f746b7afb
Variable renames, and additional debug logging.
10 years ago
0fe27651fb
Mark getConnection() that takes a username/password as deprecated. Really user's should be creating separate pools for this use-case.
10 years ago
1eebf31c05
Minor tightening of the code. Thanks, Nitin.
10 years ago
e9dc9608c9
Don't decrement timeout by deltas between intervals that may be below the resolution of nanoTime(), use an absolute base and recalculate the timeout.
10 years ago
a386815d7e
Revert "Use System.nanoTime() instead of System.currentTimeMillis() almost everywhere. System.currentTimeMillis() can go backwards!"
...
This reverts commit 01bebb5d35
10 years ago
2260cc2e5f
Make the addConnection() method a little more DRY
10 years ago
9e16052edf
Remove unnecessary throw
10 years ago
076e90b888
Test unwrap exception for Hibernate
10 years ago
81885290ed
Remove unnecessary return
10 years ago
a855c59b67
Debug log failed closes.
10 years ago
8a2e523c41
Handle edge-case where resolution of System.nanoTime() can yield zero elapsed time on busy systems.
10 years ago
01bebb5d35
Use System.nanoTime() instead of System.currentTimeMillis() almost everywhere. System.currentTimeMillis() can go backwards!
10 years ago
61e6603e2c
Refactor to remove AbstractBagEntry
10 years ago
1ebb08b532
Pass along naming exceptions, rather than wrapping them.
10 years ago
beb3fa4ebd
Minor cleanup.
10 years ago
ac7268f3c3
Minor cleanup.
10 years ago
d48b733221
Mark properties for deprecation.
10 years ago
0323a7dacb
Let PoolUtilities use its own executor. Sharing executors can lead to deadlocks (not observed, but theoretical)
10 years ago
32fa545e0f
Attempt to bound how long a connection.close() call can take to 30 seconds on JDBC41 drivers.
10 years ago
a195e7781a
Attempt to bound how long a connection.close() call can take to 30 seconds on JDBC41 drivers.
10 years ago
aa38e43195
Merged #216 proposed fix into isJava40Compliant()
10 years ago
b77806e0e4
Bump up the _first run_ of the housekeeper a bit.
10 years ago
01ea53732b
It helps to use the write value in an assertion.
10 years ago
375a8330e2
Last timing error?
10 years ago
72dd6f67ba
Remove unused import.
10 years ago
59c45c6e5f
After fillPool() change ... timing whack-a-mole game time.
10 years ago
c0047c7c46
Timing tests are a b*tch.
10 years ago
79ec2bc976
Timing tests are a b*tch.
10 years ago
63baa9ab5e
Give the pool a few beats to "settle" before asserting against counts that are arrived at asynchronously.
10 years ago
21f8f843c1
Tighter code.
10 years ago
58fccbd06c
I like this variant better.
10 years ago
cd3af22e29
Nitin thinks this is more readable. Not sure I agree, but I don't care enough to argue.
10 years ago
3d80609d6a
Nitin thinks this is more readable. Not sure I agree, but I don't care enough to argue.
10 years ago
77d6f5084a
Issue #213 allow initialization fail fast even when minimumIdle is 0.
10 years ago
3fc263b5e2
Fix a bug that was introduced in commit fb18eff that would prevent force-close warnings from being logged for connections already flagged for eviction. These warnings should still occur regardless of destined eviction.
10 years ago
cdc171f18a
JavaDoc fixes
10 years ago
c9f6d90ab2
Javadoc fixes
10 years ago
1067111883
JavaDoc fixes
10 years ago
bc42fe797a
JavaDoc fixes
10 years ago
69612f0076
Issue#201 @cowwoc merged toString() changes (largely intact)
10 years ago
fb18eff0c8
Eliminate the isForceClose member, and set bagEntry.evicted instead. We still use
...
a local isForceClose to avoid a race (where we might fail to log a broken connection).
10 years ago
7d1f0c72b5
Mainly doc fixup and method reordering.
10 years ago
590dbf1870
Add some comments. Add evicted check back to housekeeper to cover a rare race condition
...
where the connection is being released at the same time the maxLifetime eviction shedule is
running. That could have allowed the bag entry to be marked evicted, but have the check
in releaseConnection() just missing it, which would result in a maxLifetime connection sitting
in the pool for an indefinite amount of time.
10 years ago
19049c54ad
Fix #194 no need to initialize the CodaHaleShim in the JNDI factory
10 years ago
6772e5f4ca
Don't throw sun internal NotImplementedException, just throw a RunTimeException
10 years ago
fafa5050ad
Remove maxLifeTime check from getConnection() and use housekeeping scheduler instead.
10 years ago
44ecfd7f01
Fix #208 little bit cleaner impl of unwrap() and isWrapper()
10 years ago
a4c6c03ded
ConcurrentBag to common code refactor (share code between Java 6/7 and Java 8)
10 years ago
50a4de5fa4
Fix #208 all unwrap when the specified "iface" is actually a HikariDataSource
10 years ago
7e170daf15
Remove IDE warnings
10 years ago
f7415c7284
Fix #207 remove unneeded test utility class
10 years ago
62d3f4e9d9
Fix #207 opps, missed a crucial part of the fix...
10 years ago
d47d5a7b0e
Fix #207 Make PoolUtilities stateful, because using statics causes issues when two pools with different drivers and different capabilities exist in the same JVM. Pull out non-pooly things into a separate utility class.
10 years ago
7e7d124a93
Bump revision number due to large internal refactor.
10 years ago
f0b735cb02
Issue #206 Use a discard policy rather than an abort policy for the housekeeping thread, the abort policy causes unnecessary looping in addConnection() during shutdown() due to the housekeeping thread being used as the executor for the setNetworkTImeout() call. We could create another executor just for that, but it doesn't really make sense or matter.
10 years ago
5a90be9199
Make non-overridden methods final. Gets back to pre-refactor performance.
10 years ago
a6152f58a9
Merge more common code. Benchmarks are showing no negative impact so far.
10 years ago
688d1911e2
Merge more common code. HikariConfig is not a critical class, but we need to be careful about what we merge due to performance implications of inheritance and virtual dispatch that cannot be optimized by the JIT.
10 years ago
9d486abbfa
Massive structure refactor. Make IDE setup hell, but greatly improves maintainability.
10 years ago
82f98fc57e
Project restructuring experiment
10 years ago
Brett Wooldridge