diff options
| author | Wladimir J. van der Laan <laanwj@protonmail.com> | 2020-05-28 17:25:45 +0200 |
|---|---|---|
| committer | Wladimir J. van der Laan <laanwj@protonmail.com> | 2020-05-28 17:55:20 +0200 |
| commit | ea3e9e0b84c57df4110ca9e5ccced65c5bbe4611 (patch) | |
| tree | 7436aa7a3c0d8825afb25e3e7374ef87d484661b /src/txdb.h | |
| parent | 082a417abcce6a2d6e0a52ccad5cca9657cec64b (diff) | |
| parent | e8fa0a3d2025509fcddc59fc618e91371542cf87 (diff) | |
| download | bitcoin-ea3e9e0b84c57df4110ca9e5ccced65c5bbe4611.tar.xz | |
Merge #18700: Fix locking on WSL using flock instead of fcntl
e8fa0a3d2025509fcddc59fc618e91371542cf87 Fix WSL file locking by using flock instead of fcntl (Samuel Dobson)
Pull request description:
Fixes #18622
A bug in WSL means that fcntl does not exclusively lock files, allowing multiple instances of bitcoin to use the same datadir. If we instead use flock, it works correctly. Passes Travis, but testing on some OS variety would be sensible.
From what I can tell, flock and fcntl don't work with each other on linux, so it would still be possible to run a node with this code change and a node before it with the same datadir (this isn't true for Mac/FreeBSD). flock also doesn't support NFS on MacOS and linux<2.6.12 while fcntl did. See here for example: https://gavv.github.io/articles/file-locks/
If changing to flock for all systems is inadvisable, it would also be possible to just detect WSL and use flock when on that platform to avoid the bug.
ACKs for top commit:
laanwj:
Code review ACK e8fa0a3d2025509fcddc59fc618e91371542cf87
Tree-SHA512: ca1009e171970101f1dc2332c5e998717aee00eebc80bb586b826927a74bd0d4c94712e46d1396821bc30533d76deac391b6e1c406c406865661f57fa062c702
Diffstat (limited to 'src/txdb.h')
0 files changed, 0 insertions, 0 deletions