fix(driver): load_replace_candidate is None, not a stub that raises (NetOrk/netork#2503)
driver / lint (pull_request) Successful in 19s
driver / standard (pull_request) Failing after 27s
driver / types (pull_request) Successful in 3m34s
driver / test (3.13) (pull_request) Successful in 3m12s
driver / test (3.12) (pull_request) Successful in 3m23s
driver / lowest (pull_request) Successful in 35s
driver / build (pull_request) Successful in 35s
driver / security (pull_request) Successful in 1m15s
CI / driver (pull_request) Failing after 4m54s
driver / publish (pull_request) Skipped

The driver cannot replace a device's configuration, and said so with a
load_replace_candidate that raised NotImplementedError. A method that only
raises still looks like a replace to anyone who asks whether the driver has
one: netOrk offered a config restore for these devices that could only fail.

None is how a driver switches off a method, the convention netOrk's
capability checks and napalm-device-types follow, so asking now says no.
Nothing in the driver calls it.
This commit is contained in:
Christian Manivong committed 2026-11-13 21:45:23 +01:00
1 parent 5f6595d160
commit c9a9d532aa
4 files changed
+25 -3

No files matched your search

+4
View File
@@ -21,6 +21,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
`NotImplementedError`.
### Changed
- `load_replace_candidate` is `None` instead of a method that raises
`NotImplementedError`, so a caller that asks whether the driver can replace a
configuration gets no. netOrk offered a config restore for these devices that
could only fail (NetOrk/netork#2503).
- On an rpm host a package's version carries its epoch where it has one
(`1:6.2.0-13.el9`), as on a dpkg host. Oracle's and Red Hat's fixes name it, and without
it an epoch-1 fix read as newer than every installed build (NetOrk/netork#2394).
+1 -1
View File
@@ -91,7 +91,7 @@ with Driver(
| `get_config()` | ✅ | Returns `ip addr` + `ip route` output |
| `ping()` | ✅ | Executes `ping` on the remote host |
| `load_merge_candidate()` | ❌ | Not applicable for generic Linux |
| `load_replace_candidate()` | ❌ | Not applicable for generic Linux |
| `load_replace_candidate()` | ❌ | Not applicable for generic Linux; `None`, so a caller can tell |
| `compare_config()` | ❌ | Not applicable for generic Linux |
| `commit_config()` | ❌ | Not applicable for generic Linux |
| `discard_config()` | ❌ | Not applicable for generic Linux |
+3 -2
View File
@@ -1009,8 +1009,9 @@ class LinuxDriver(
def load_merge_candidate(self, filename: str = None, config: str = None) -> None: # type: ignore[override]
raise NotImplementedError("Config management is not supported for Linux hosts")
def load_replace_candidate(self, filename: str = None, config: str = None) -> None: # type: ignore[override]
raise NotImplementedError("Config management is not supported for Linux hosts")
#: No config replace: switched off rather than a stub that raises, so asking
#: whether the driver has one says no (NetOrk/netork#2503).
load_replace_candidate = None # type: ignore[assignment]
def compare_config(self) -> str:
raise NotImplementedError("Config management is not supported for Linux hosts")
+17
View File
@@ -0,0 +1,17 @@
"""The driver cannot replace a device's configuration, and says so by not having the method.
A ``load_replace_candidate`` that only raises looks like a replace to anyone who asks
whether the driver has one: netOrk offered a config restore for these devices that
could only fail (NetOrk/netork#2503). ``None`` is how a driver switches off a method
it would otherwise inherit, the convention of napalm-device-types and netOrk's
capability checks.
"""
from __future__ import annotations
from napalm_linux import LinuxDriver
def test_the_driver_has_no_config_replace():
assert LinuxDriver.load_replace_candidate is None
assert not callable(getattr(LinuxDriver, "load_replace_candidate", None))