`Scripts.razor` rendered a hash prefix as `@s.SourceHash[..12]…`. `SourceHash`
is a plain nvarchar(64) with no length floor, so any value shorter than 12
characters threw ArgumentOutOfRangeException out of BuildRenderTree — unhandled
in Blazor, which takes the WHOLE page to an HTTP 500, not just the offending
row. Live-reproduced on docker-dev, where two Script rows carry `h1` and
`h-abs-hr200`: /scripts was a hard 500.
New `Components/Shared/DisplayText.Abbreviate(string?, int)`:
- null/blank renders the Admin UI's em-dash placeholder rather than throwing;
- a value already within budget renders verbatim with NO ellipsis, so the
display never implies text that isn't there (the old markup appended "…"
unconditionally, outside the slice — `h1` would have shown as `h1…`);
- a maxLength < 1 is clamped, so a display bug cannot become a page crash.
Applied to every unguarded range-operator slice over a DB-sourced string found
by sweeping the AdminUI for `[..N]`:
Scripts.razor SourceHash (the reported crash)
Certificates.razor Thumbprint (read off the on-disk store)
Deployments.razor x2 RevisionHash
Clusters/ClusterOverview.razor RevisionHash
The other `[..N]` hits are safe and untouched: Guid.ToString("N") slices
(always 32 chars) and the `Length > 60 ? [..60]` ternaries that self-guard.
NOTE — the issue text blamed the docker-dev seed; that was wrong and is
corrected on #504. `seed-clusters.sql` inserts only ServerCluster, ClusterNode
and LdapGroupRoleMapping. The short-hash rows came from earlier live-gate
authoring. A freshly seeded rig is fine; the trigger is any Script row that did
not go through ScriptEdit / UnsTreeService.HashSource — REST-API authored,
hand-inserted, or migrated in.
Tests: 14 new in DisplayTextTests, covering the short/null/blank/clamped cases
and a never-throws sweep over every value x budget combination. AdminUI suite
739/739.
Live-verified on docker-dev (AdminUI has no bUnit): /scripts 500 -> 200 across
both central nodes, rendering `hash=h1` and `hash=h-abs-hr200` in full;
/deployments still truncates 64-char hashes at 12 with the ellipsis;
/certificates and /clusters/MAIN unchanged.
Same single-page edit-or-create pattern as batch 1, applied to the
foundational topology entities. After this batch the whole hierarchy
(cluster → nodes → UNS areas → UNS lines → namespaces → drivers) is
fully editable through the UI.
- ClusterEdit.razor /clusters/{id}/edit
Update + delete for an existing cluster. NodeCount stays coupled to
RedundancyMode (None→1, Warm/Hot→2). ModifiedBy taken from
AuthenticationStateProvider.
- NodeEdit.razor /clusters/{id}/nodes/{new|nodeId}
Full ClusterNode CRUD. ApplicationUri uniqueness is enforced by EF
index; ServiceLevelBase defaults to 200 (primary preference) on
create; per-node DriverConfigOverridesJson validated as JSON.
- UnsAreaEdit.razor /clusters/{id}/uns/areas/{new|id}
- UnsLineEdit.razor /clusters/{id}/uns/lines/{new|id}
UNS structure CRUD; Lines pick their parent Area from a select that
loads the cluster's areas.
List pages updated:
- ClusterOverview now shows an "Edit cluster" button + a "New node"
action on the nodes panel + per-row Edit buttons.
- ClusterUns gains New/Edit affordances for both Areas and Lines.
All 9 integration tests still green; no regressions.
Per Q3 of the rebuild plan, each v1 ClusterDetail tab becomes a separate
route under /clusters/{id}/<tab>. This batch adds read-only table views
for the six core config entity types; live-edit forms with RowVersion
concurrency land in Phase C.2 once the read-view shape is reviewed.
- ClusterEquipment /clusters/{id}/equipment — joins via DriverInstance
so the cluster scope works
- ClusterUns /clusters/{id}/uns — Areas + Lines tables
- ClusterNamespaces /clusters/{id}/namespaces — Kind + URI + Enabled chip
- ClusterDrivers /clusters/{id}/drivers — collapsed list with JSON
config expandable per Q1
(typed editors deferred)
- ClusterTags /clusters/{id}/tags — first 200 by name + filter
- ClusterAcls /clusters/{id}/acls — LDAP group + scope +
NodePermissions bits
Shared ClusterNav.razor extracted; ClusterOverview + ClusterRedundancy
updated to use it. _Imports.razor adds Components.Shared so the shared
nav is in scope across pages.
- ClustersList (/clusters) — table view, row-click opens detail
- NewCluster (/clusters/new) — EditForm with DataAnnotations; redundancy
mode + node-count coupling enforced client-side (None→1, Warm/Hot→2);
CreatedBy taken from AuthenticationStateProvider
- ClusterOverview (/clusters/{id}) — cluster details + last-deployment
badge + node list. Per Q3, the legacy 10-tab monolith is split into
separate routes; this page hosts the Overview "tab" as its primary slot
- ClusterRedundancy (/clusters/{id}/redundancy) — static ServiceLevelBase
config view; live ServiceLevel comes via RedundancyStateActor DPS topic
(deferred to its own follow-up once the SignalR bridge lands)
The other 8 v1 cluster tabs (Equipment, UNS, Namespaces, Drivers, Tags,
ACLs, ScriptedAlarms, Scripts, Audit) land in Phase C/D.
Joseph Doherty