It seemed to create two CWalletDB objects that both grab the database lock.
(note: based on v0.7.1)
It seemed to create two CWalletDB objects that both grab the database lock.
(note: based on v0.7.1)
It seemed to create two CWalletDB objects that both grab the
database lock.
ACK
I'm probably overlooking something, but if the lock is grabbed twice in the same thread, how can that cause a deadlock with recursive mutexes?
It would seem the BDB locks are not recursive.
Right,it wasn't clear to me that this concerns the bdb locks, thanks. ACK ofc.
@gavinandresen Do you think this warrants a 0.7.2?
Yes, I think this is serious enough to warrant a 0.7.2.
I have 0.7.x ready-to-tag as soon as this gets pulled in to master.
Other fixes included:
@luke-jr I would not include https://github.com/bitcoin/bitcoin/commit/6b3783a as this needs the bitcoinstrings.cpp update (#1999) and a new master file on Transifex, which is what I'm waiting for to start translating the german stuff, to have the changes in for all languages. It's clearly for 0.8 (and already merged in master) and non-critical + needs time to mature.
NACK on these, as I don't think they pass the risk/benefit test:
- 128eefa https://github.com/bitcoin/bitcoin/commit/128eefa Qt: small header changes / fixes
- 6b3783a https://github.com/bitcoin/bitcoin/commit/6b3783a fix some double-spaces in strings
- 0f8a647 https://github.com/bitcoin/bitcoin/commit/0f8a647 don't use memset() in privacy/security relevant code parts
Gavin Andresen
@gavinandresen I disagree here, as I think https://github.com/bitcoin/bitcoin/commit/0f8a647 is quite an important fix / change.