Rendering the SSO buttons as POST forms (#23042) made every SSO login a form
submission which NetBox answers with a redirect to the identity provider.
Chromium-based browsers evaluate the CSP form-action directive against every hop
in a form submission's redirect chain, so a deployment which serves NetBox with
`form-action 'self'` blocks that redirect and the button silently does nothing.
Add SocialAuthBeginView, which wraps python-social-auth's begin view and returns
the identity provider's URL as JSON to clients which request it. The login page
now submits the form via fetch() and assigns window.location, which form-action
does not govern. The upstream view is reused as-is, so CSRF protection, the
callback URL, and the session state recorded for the identity provider are
unchanged; clients which do not request JSON (a browser without JavaScript, or a
backend which renders an HTML form rather than redirecting) receive the
unmodified response as before.
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
The existing tests assert on filter results, which stay correct for ASCII values
even when the collation is never applied. Add a test which asserts on the
lookup's own compiled output, so that the mechanism failing open is caught
rather than passing silently.
Explain why the placeholder is compared literally: a field declaring its own
get_placeholder() compiles to something other than '%s', and splicing a COLLATE
clause into that is not safe, so any other right-hand side is left alone.
The documentation note described the folding as specific to the German
eszett. It is the common example rather than the rule: the collation treats a
character as equivalent to the sequence it expands to in upper case, which also
covers ligatures. Note too that a case-insensitive exact match on a collated
field may now return more than one object.
Django's PostgreSQL backend compiles icontains, iexact, istartswith and
iendswith as UPPER(col::text) LIKE UPPER(%s). UPPER() folds according to the
collation of its argument, and the two sides do not share one: the column folds
under its own collation while the parameter folds under the database default.
For a column using natural_sort, UPPER('ß') is 'SS' on the left and 'ß' on the
right, so searching for 'ß' matched nothing. This affects the name field of most
models and all four case-insensitive lookups, which are also exposed through the
REST API as __ic, __ie, __isw and __iew.
Apply the column's collation to the parameter as well, inside the UPPER() call,
so that both sides fold the same way. Matching on those fields becomes
bidirectional, so 'Strasse' finds 'Straße' and vice versa. Fields without the
collation are unchanged.
The lookups only collate a bare column compared against a simple value. An
expression which already carries an explicit collation, such as Collate() or
CollateAsChar(), would otherwise raise a collation mismatch error.
Validate REST script input before enqueueing jobs. Resolve ObjectVar
IDs to model instances and MultiObjectVar IDs to querysets, returning
HTTP 400 with errors nested under data when validation fails.
Share form preparation between the API and UI, including multi-value
defaults, while keeping validation out of the job runner to preserve
other execution paths. Exclude only known execution fields from script
data and prevent _notifications from leaking into CLI script input.
Document the REST compatibility changes, including required-field
validation and discarded undeclared keys. Add regression coverage for
object resolution, defaults, validation errors, and execution options.
Co-authored-by: Martin Burggraf <martin.burggraf@netclusive.com>
Document the database permissions required to install the ltree extension
during an upgrade to NetBox v4.7. Clarify that installations following the
standard PostgreSQL setup already satisfy this requirement through database
ownership.
Provide commands for granting the database-level CREATE privilege where
needed, along with an administrator pre-installation option for deployments
using a restricted NetBox database role.
Co-authored-by: Martin Hauser <mhauser@netboxlabs.com>
- Document that ModuleType.to_yaml() exports module_bay_types by name but
the field isn't currently importable back through it (no ModuleTypeImportForm
field survived the CSV-import revert).
- modulebaytemplate.md's note covered only the device-type-parented import
path; ModuleBayTemplateImportForm is registered for both DeviceTypeImportView
and ModuleTypeImportView, scoping to whichever parent type's manufacturer
applies. Reworded to cover both, and added the "rejected rather than
resolved" clause for a name matching only some other manufacturer's type.
- Clarified clean_module_bay_types()'s docstring: the "never a cross-manufacturer
collision" guarantee holds only because ModularComponentTemplateModel.clean()
rejects a template with neither device_type nor module_type before this
method's result would ever be saved.
- Fixed a test docstring overstating symmetry between its two comparison arms.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Per review, ditch dedupe_module_bay_types_by_manufacturer() and any import
logic that resolves module_bay_types by name alone across manufacturers.
ModuleBayType's unique constraint is (manufacturer, name), not name alone,
so resolving a bare name against an unscoped, potentially cross-manufacturer
pool -- guessing via a preference order, rejecting only on a genuine tie --
is not a reliable way to identify a specific object. CSV import forms have
no way to qualify an M2M reference beyond a bare name, so module_bay_types
is no longer exposed there at all (ModuleTypeImportForm, ModuleBayImportForm
in bulk_import.py): it's acceptable not to support this rather than resolve
it unreliably. This also reverts the netbox/tables/columns.py export_transform
API addition and the three tables' use of it, which existed only to make the
CSV round trip work.
The one import path that survives is ModuleBayTemplateImportForm (the YAML
device/module type "Import Components" flow), because it can reliably scope
module_bay_types' queryset to the parent device/module type's own
manufacturer plus global (manufacturer-less) types *before* resolving by
name -- so a name collision is never cross-manufacturer, only "this
manufacturer's own type vs. a global one of the same name," which
ModuleBayType's own uniqueness constraint makes unambiguous. A name matching
only some other manufacturer's type doesn't resolve at all, rather than
being coerced to an arbitrary guess.
Kept: the ModuleBayTemplateImportForm.enabled field/clean_enabled() fix
(default=True was previously lost on YAML re-import; unrelated to the
above), and to_yaml()'s export of module_bay_types on both ModuleType and
ModuleBayTemplate, plus the export_yaml() prefetch optimizations -- none of
these involve resolving an object's identity from an ambiguous attribute.
Trimmed the model docs to match: the modulebay.md and moduletype.md
paragraphs described capabilities (CSV import, cross-manufacturer YAML
import) that no longer exist and are removed; modulebaytemplate.md's note
is rewritten to describe the actual (manufacturer-or-global-scoped)
resolution behavior.
Allow channel subinterfaces to retain a specific physical interface type
and rename conventionally named children when their parent is renamed.
Keep mirrored cable and path state consistent when channel bindings
change, avoid unnecessary path rebuilds, and apply the same rename
behavior to interface templates.
Distinguish absent values from malformed or unwalkable paths during
condition evaluation.
Preserve valid empty-list traversal, support changes in snapshot shape,
and reject snapshot attributes which are missing from both snapshots.
Normalize absent job payloads and ensure conditioned rules fail closed
when no payload is available. Add regression coverage and streamline the
related documentation and comments.
- ModuleBayType.__str__() includes the manufacturer (e.g. "Cisco SFP28"),
but the three module_bay_types ManyToManyColumn declarations had no
transform, so django-tables2 defaulted to str() for CSV export while
the import forms resolve by name alone. Verified directly: exporting
a manufacturer-scoped bay type produced "Cisco SFP28", which then
failed to re-import with "Object not found: Cisco SFP28" -- broken
for exactly the case (manufacturer-scoped types) the preference/
ambiguity machinery exists to serve. Set transform=lambda obj: obj.name
on all three columns to match to_yaml(), and rewrote the existing
round-trip test to use a manufacturer-scoped bay type instead of a
global one, which is the only case that exercised str().
- The three ambiguity-refusal tests asserted only that the field
errored, which a plain invalid_choice (e.g. from a queryset that
excluded both candidates) would also satisfy -- masking a regression
of the manufacturer scoping removed two commits ago. Tightened each to
assert the error names both competing manufacturers.
- Corrected modulebay.md, which still described module_bay_types
resolution as "scoped to" the device's manufacturer -- the behavior
the prior commit removed as a bug; it's a preference, not a scope.
- Trimmed comments and docstrings introduced across this branch to a
more proportionate length.
Deliberately out of scope for this PR (tracked as follow-up
considerations, not fixed here): an escape hatch for a bay type name
that's genuinely ambiguous across manufacturers with no local match
(would require a new wire-format convention), and ModuleType.to_yaml()
not exporting a module-bays section at all (a separate, pre-existing
asymmetry, larger than this PR's scope).
- CSVModelMultipleChoiceField.clean() split on a bare comma with no
whitespace stripping, but ManyToManyColumn's default CSV export
separator is ", " (comma + space) -- so re-importing NetBox's own CSV
export of any multi-value column using this field (module_bay_types
among others, since this is a shared utility field) failed with
"Object not found: <value>" on every value after the first. Verified
directly against ModuleTypeTable's actual export value before fixing.
Also cast to str() before splitting: a YAML-bound caller (as opposed to
a CSV cell, always a string) can pass a non-string scalar, which
previously raised an unhandled AttributeError instead of a form error.
- Docs for module bay type resolution still described the pre-a3b5e4b
fallback ("then any remaining candidate"); updated to describe the
refusal behavior that replaced it. Added a matching note to
modulebay.md, which had none.
- dedupe_module_bay_types_by_manufacturer() collapses candidates by pk
within each name group before computing preference, so a caller
passing a duplicate row in a raw list -- the signature accepts "an
iterable," not just a queryset -- can't manufacture a same-manufacturer
tie that would then crash on None.manufacturer.name. Unreachable via
the three current callers today (each resolves from a queryset,
which can't contain a row twice), but cheap to make the helper safe
standalone.
- Fixed a stale test docstring contrasting the two import forms' field
types by a distinction (plain vs. CSV multiple-choice field) that no
longer exists since both were aligned to CSVModelMultipleChoiceField.
- Added ambiguity-refusal coverage at the other two call sites
(ModuleBayTemplateImportForm, ModuleBayImportForm) -- previously only
ModuleTypeImportForm was covered for this path.
Also found independently while verifying the above: ModuleTypeListView
.export_yaml() prefetched modulebaytemplates__module_bay_types, but
ModuleType.to_yaml() -- unlike DeviceType.to_yaml() -- never reads
self.modulebaytemplates at all (a separate, pre-existing, out-of-scope
gap: ModuleType.to_yaml() doesn't export a nested module-bays section).
That prefetch was dead weight, adding a query with no corresponding
saving. Removed it, and with it the now-meaningless "bay count doesn't
affect query count" test (nothing in ModuleType.to_yaml() ever varied
with bay count to begin with), replacing it with an exact-delta
assertion isolating what the one relevant prefetch (module_bay_types
on the module type itself) actually saves.
Pin `twine` and `packaging` versions in build job to match bundled
versions in `gh-action-pypi-publish` v1.14.2.
Enforce Core Metadata 2.4 in wheel and sdist targets with verification
in validation scripts.
Pin `twine` and `packaging` versions in build job to match bundled
versions in `gh-action-pypi-publish` v1.14.2.
Enforce Core Metadata 2.4 in wheel and sdist targets with verification
in validation scripts.
Fixes#22903
The port-mappings CSV column expanded only comma-separated individual
protocol/port pairs, while the edit form's port field already accepted
hyphen ranges (e.g. tcp/8000-8010). Route the CSV column through the same
expand_port_mapping() helper so both entry paths accept identical port
syntax. Parsing uses the shared split_port_mapping() helper, and the
blank-protocol error is worded to fit every entry path rather than only the
form widget's dropdown.
The manufacturer-or-null queryset scoping added to disambiguate a name shared
by a global and a manufacturer-scoped ModuleBayType went further than
intended: it also excluded a *different* manufacturer's bay type entirely.
The UI (ModuleTypeForm/ModuleBayTemplateForm) and REST API place no such
restriction -- a third-party module may legitimately declare compatibility
with another manufacturer's proprietary bay type. Confirmed the regression
directly: creating that assignment via ModuleTypeForm succeeds, but
exporting it and re-importing the same YAML failed with
"Object not found: SFP28", making valid existing data unimportable -- worse
than the bug this feature exists to fix.
Remove the queryset scoping entirely and instead make
dedupe_module_bay_types_by_manufacturer() manufacturer-aware: given the
target manufacturer, it now prefers (in order) an exact match, then a global
type, then any remaining candidate, resolved from an unscoped queryset in
clean() rather than a sibling clean_<field>() mutating the field's queryset.
This also drops the Meta.fields-ordering dependency those methods required.
Also, from the same review round:
- Test asserting Django's literal English error string now asserts the
error code instead, so it survives wording changes/translation.
- The prefetch query-count test moved from test_models.py (which doesn't
otherwise touch views) to test_views.py, and strengthened from "prefetch
saves at least one query" to "query count is constant regardless of bay
count" -- the actual invariant. Added equivalent coverage for
ModuleTypeListView, which the prior version didn't test at all.
- Corrected the export_yaml() prefetch comments to not imply the other
~11 relations to_yaml() touches are also covered -- they aren't, and
weren't before this feature either.
- Updated the model docs to describe the new (permissive, cross-manufacturer
allowed) behavior instead of the old (restrictive) one they described a
commit ago.
Adds regression tests importing a bay type belonging to a different
manufacturer than the importing device/module type, through both
ModuleBayTemplateImportForm and ModuleTypeImportForm.
- clean_module_bay_types()'s two ValidationErrors were unreachable:
ModelMultipleChoiceField.clean() already raises before the clean_<name>
hook runs on a non-list or an unresolvable name, per Django's
BaseForm._clean_fields(). Simplify to dedupe from cleaned_data (already
scoped and validated) via a shared dedupe_module_bay_types_by_manufacturer()
helper in dcim/utils.py, used by both ModuleBayTemplateImportForm and the
new ModuleTypeImportForm.module_bay_types below. This also drops the
self.data access that ignored the form prefix, broke on a QueryDict, and
re-queried once per name.
- ModuleType.module_bay_types (the module's own side of the bay/module
compatibility intersection) was still missing from the YAML round trip.
Add it to ModuleType.to_yaml() and ModuleTypeImportForm, mirroring
ModuleBayTemplateImportForm's manufacturer-scoping and dedup.
- ModuleBayTemplate.to_yaml() emitted enabled but the import form didn't
accept it, so it silently reset to False (not the model's default=True)
on any dict-bound re-import. Add it with the same
clean_enabled()-defaults-to-True pattern already used by
ModuleBayImportForm's CSV import.
- Prefetch module_bay_types in DeviceTypeListView/ModuleTypeListView's
export_yaml() so bulk YAML export doesn't add one query per module bay
template across the exported queryset.
- Document the manufacturer-preference precedence rule in the model docs,
since export emits a bare name and import can resolve a colliding one to
either a global or manufacturer-specific type.
Adds regression tests for the module_type-scoped path, the enabled
round trip, an export/import round trip, export ordering, the new
ModuleTypeImportForm coverage, and the prefetch fix.
* Add support for liquid cooling components
* Include sample of offending components when module move is disallowed
* Use settings.BULK_UPDATE_CHUNK_SIZE for batch_size
* Adopt review feedback
Resolves all conflicts between main and feature for #22896. Notable
resolutions:
- dcim/signals.py, dcim/tests/test_signals.py: main's cache_presave_scope_fields
/ sync_cached_scope_fields addition is fully superseded by feature's
PostgreSQL-trigger-based denormalization (confirmed via feature's own
migration docstrings); dropped in favor of feature's existing approach.
Both files now match feature's originals exactly.
- netbox/tables/columns.py: combined main's generic get_ordering_annotation()
protocol with feature's nulls_first-aware order() override. These two
mechanisms cannot both apply to the same column (django-tables2 negates an
entire order_by tuple uniformly on direction toggle, so a fixed nulls_first
placement and multi-column sort composition are mutually exclusive for one
column) -- preserved nulls_first (existing, wired through forms/API/GraphQL)
and removed main's two composition-only tests for CustomFieldColumn. See the
comment on CustomFieldColumn.order() for full reasoning.
- extras/customfields.py, extras/graphql/mixins.py: combined main's
request-cache optimization and has_key-scoped batch updates with feature's
resolve_selection_value() (shared select-field label resolution between
REST and GraphQL).
- extras/events.py, extras/event_rules.py: main's "Honor Script defaults when
triggered by Event Rules" (#22852) fix was written against the old inline
action-type dispatch, which feature had already replaced with a pluggable
action-provider registry (#22770). Re-applied the same two-line fix
(notifications/job_timeout) inside ScriptAction.enqueue() in event_rules.py
instead.
- utilities/jinja2.py: fixed a config-attribute name mismatch the raw merge
would have introduced (main's JINJA2_FILTERS vs feature's renamed
JINJA_FILTERS) by updating the shared _jinja2_filters() helper.
- ipam/migrations/: renumbered main's 0094_ipaddress_host_index to 0096 and
added a merge migration, since main and feature had each independently
added a migration numbered 0094.
- dcim/tests/query_counts.json: regenerated via UPDATE_QUERY_COUNTS=1 against
the merged codebase rather than hand-merging counts.
Verified: manage.py check clean, full migration graph applies cleanly from
scratch, ruff clean, and full test suites pass for dcim, ipam, netbox, extras,
circuits, vpn, wireless, tenancy, virtualization, core, users, and account
(fresh databases, no state carried over between runs).
Normalize RQ timeout values before validating global and per-webhook
timeouts, including duration strings and RQ's default and unlimited values.
Improve timeout logging and visibility in the UI and documentation, raise
the default webhook timeout to 60 seconds, and add coverage for the new
validation and filtering behavior.
Jeremy Stretch
Jason Novinger
Arthur Hanson
Martin Hauser
Mohamed Hossam
bctiemann
Brian Tiemann