[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#1115317: Technical Committee resolution on /var/lock and systemd



On 10/8/25 7:09 AM, Josh Triplett wrote:
Simon Richter wrote:
On 10/8/25 6:40 AM, Josh Triplett wrote:

- First, port everything using `/var/lock` or `/run/lock` to lock
   something that has an obvious `flock` target (e.g. a serial
   device), one-by-one, to use `flock` *and*` a lockfile, silently
   ignoring permission errors iff it successfully acquired the
   `flock`.

We can do this on Linux-with-udev, because /dev is guaranteed to
support flock(). This might break on some embedded devices with
pre-created device nodes (for faster boot), and on NFS-root, but
everyone using one of these has left Debian years ago.

Right.

- Second, when everything has been ported *and* we've been through
   some transition period (using package dependencies in Debian and
   allowing some time for third-party sotware), we let software start
   deleting the lockfile code.

The actual code will probably need to remain because we can't make
decisions for non-Linux here. We might want to disable it.

"let software start deleting", not "force software to start deleting".
It's up to individual programs how long they want to keep the non-flock
lockfile compat code, and what systems other than Linux they expect to
be compatible with.

I keep wondering if it's worth abstracting this into a library - instead of everyone trying to get it right themselves. But then something would need to come forward and implement that.

Kind regards
Philipp Kern


Reply to: