Use database locking (flock) for zsh #199
Conversation
|
@rupa ping. Could you please review and merge this? I've been using it for two months and so far had no races and DB corruptions. |
|
@rupa ping |
1 similar comment
|
@rupa ping |
|
Yes please! Would love to see this get merged. |
|
@rupa gentle ping |
|
This would be very useful.. |
|
As this PR is coming up on two years, I'm closing it. If there's anything I can do to make it land, I'd be happy to fix. For example, if the use of |
|
sorry my friend, I haven't have a lot of bandwidth, I don't use zsh, and I am definitely slow on changes with this project for a variety of reasons. This is an issue though, and i'd prefer it stay open. Thanks, and apologies |
This commit ensures the database is locked before reading and updating when used in zsh. We load the zsh/system module which provides the zsystem flock command to perform advisory file locking. A separate code path is needed when updating the database entry because replacing the file (via mv) would cause any process waiting for a lock to fail. Here we simply clobber the database with the contents of the tempfile.
|
@rupa that's quite alright, I understand 😄. Rebased on master to get rid of the conflicts. |
This PR adds support for locking the database (via advisory file locking) during update when
zshis used as shell. This is done viazsystem flockthat is available in thezsh/systemmodule. The lock is acquired before the database is read and kept until the (new) database has been written and the command has exited.A separate code path is needed when updating the database entry because replacing the file (via mv) would cause any process waiting for a lock to fail. Here we simply clobber the database with the contents of the tempfile.
This is a partial fix for #198, more specifically, this only fixes the problem on
zsh.// ping @balta2ar