Closed afbjorklund closed 2 months ago
LGTM @baude @ashley-cui PTAL
Does this count as a breaking change, should we deprecate this first?
It was a breaking change already in 5.0, when the flag got disconnected...
commit 9bb191df51c7f2d242835e2af0bfa19d781b2256 ("70 files changed, 2372 insertions(+), 2473 deletions(-)")
But there were never any options for the flag anyway, so it was a no-op.
var volumeType string
switch opts.VolumeDriver {
// "" is the default volume driver
case "virtfs", "":
volumeType = VolumeTypeVirtfs
default:
return fmt.Errorf("unknown volume driver: %s", opts.VolumeDriver)
}
But if anyone is actually using it, it would break their command line.
Can probably remove it as the same time as changing to virtiofs, though? (5.2)
Otherwise it would just be a lie, since the actual implementation has changed.
Maybe mark the flag as hidden for a few releases (print a warning on use) and remove in ~5.4 timeframe later this year?
Maybe mark the flag as hidden for a few releases (print a warning on use) and remove in ~5.4 timeframe later this year?
Wouldn't you have to change the value of it then? It would be weird to continue using --volume-driver=virtfs
in 5.2
EDIT: Oh, you mean leave it as a no-op (like it is now) but just hide it from the init flags and the documentation
@mheon @ashley-cui PTANL?
I marked it as "deprecated" instead, even though it was removed in Podman 5.0 already.
It is still hidden (and the docs removed), but now you get a warning if you try to use it:
Flag --volume-driver has been deprecated, will be ignored
/approve /lgtm
[APPROVALNOTIFIER] This PR is APPROVED
This pull-request has been approved by: afbjorklund, rhatdan
The full list of commands accepted by this bot can be found here.
The pull request process is described here
The driver is now hardcoded again, and there can only be one type of mounts at a time (which one changes over time)
Revert "Make it possible to select the volume driver" This reverts commit 6630e5cf66cf76aefcfe9caebe5df4f37dd0bdd5.
The driver is now always virtiofs(d).
Previously it was always virtfs (9p).
Does this PR introduce a user-facing change?
/cc @baude
The flag was introduced, so that users could choose between sshfs and virtfs drivers:
Mostly because some operating systems do not support 9p filesystems, such as RHEL
The real name should have been "reverse-sshfs", since it is calling back (to sftp on host)