| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
| |
Those would, if uncaught, exit run and leave the waiter to wait
indefinitely for the number of active jobs to reach 0
|
| |
|
|
| |
Update copyright year to 2020
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Avoids a DB error (leading to an assert) where a thread uses
a read txn previously created with an environment that was
since closed and reopened. While this usually works since
BlockchainLMDB renews txns if it detects the environment has
changed, this will not work if objects end up being allocated
at the same address as the previous instance, leading to stale
data usage.
Thanks hyc for the LMDB debugging.
|
| | |
|
| |
|
|
|
| |
Decrease the number of worker threads by one to account
for the fact the calling thread acts as a worker thread now
|
| |
|
|
|
|
| |
use unsigned int to avoid having to range check negative numbers,
use const where possible, don't needlessly create empty objects,
use std::move where possible
|
| |
|
|
|
|
|
|
|
|
| |
If an exception is thrown, it is ignored. While this may hide
a bug, this should only be system exceptions in boost, which
is pretty unlikely. Morever, wait should be called manually
before the dtor anyway. Add an error message if the dtor has
to wait in case some such cases creep in so they get fixed.
Coverity 182538
|
| | |
|
| | |
|
| |
|
|
| |
Create capacity for 2x max, but lie about it
|
|
|
Instead of constantly creating and destroying threads
|
