Commit Graph
3 Commits
Author SHA1 Message Date
romanettandClaude Opus 5 1af191ec26 Drop the .Common suffix from the three GDS projects (#4438)
# Description

Every other companion specification in the stack ships as `<Family>`,
`<Family>.Client` and `<Family>.Server` — `Opc.Ua.Di`, `Opc.Ua.ISA95`,
`Opc.Ua.Positioning`, `Opc.Ua.Robotics`, `Opc.Ua.Vision`,
`Opc.Ua.WotCon`, `Opc.Ua.XRegistry`, and `Opc.Ua.Lds.Server` without a
family package. GDS was the one exception, carrying a `.Common` suffix
that named nothing.

| Before | After |
| --- | --- |
| `src/Opc.Ua.Gds.Common` | `src/Opc.Ua.Gds` |
| `src/Opc.Ua.Gds.Client.Common` | `src/Opc.Ua.Gds.Client` |
| `src/Opc.Ua.Gds.Server.Common` | `src/Opc.Ua.Gds.Server` |

## No source changes were needed

`RootNamespace` was already the final name in all three projects — the
namespaces have always been `Opc.Ua.Gds`, `Opc.Ua.Gds.Client` and
`Opc.Ua.Gds.Server`, and only `AssemblyName` and `PackageId` disagreed.
Nothing that compiles against these assemblies changes a `using`
directive; the diff outside the moves is entirely ids, paths and prose.

Directories were moved with `git mv`, and every one of the 73 file moves
is detected as a rename at 100% similarity (bar the handful that also
had a name in their text), so `git log --follow` and blame survive.

## Consumer impact

The package ids change, which is visible to anyone consuming the NuGet
packages:

| 1.5.378 package | 2.0 package |
| --- | --- |
| `OPCFoundation.NetStandard.Opc.Ua.Gds.Common` |
`OPCFoundation.NetStandard.Opc.Ua.Gds` |
| `OPCFoundation.NetStandard.Opc.Ua.Gds.Client.Common` |
`OPCFoundation.NetStandard.Opc.Ua.Gds.Client` |
| `OPCFoundation.NetStandard.Opc.Ua.Gds.Server.Common` |
`OPCFoundation.NetStandard.Opc.Ua.Gds.Server` |

`docs/migrate/2.0.x/packages.md` gains a renamed-packages section — it
is the document that undertakes to cover NuGet renames, and had no GDS
entry — and the migration skill's package table gains the two rows a
1.5.378 consumer needs. That is the only hand-written content in this
PR; everything else is mechanical.

Assembly names follow the package ids, so the section also notes that
binding redirects, trimming descriptors and signing manifests naming the
files directly have to be updated.

## Build infrastructure

Everything naming the projects or their output moves with them:

- `UA.slnx` and `tools/MigrationAnalyzer.slnx` (both re-sorted — the
project lists are alphabetical, and `Gds` now sorts before
`Gds.Client`).
- `.azurepipelines/expected-packages.txt` — 3 ids, also re-sorted for
the same reason.
- `.azurepipelines/signlistDebug.txt` and `signlistRelease.txt` — 18
paths each. Their internal ordering is left alone: that group was
already unsorted, and reordering a signing manifest is diff noise.
- `samples/Reference/ConsoleReferenceServer/Dockerfile`, the two
metapackage nuspecs, the analyzer nuspec's package dependencies, and 8
`.csproj` references across tests, samples and tools.

`preview-pack.slnx` is generated from `UA.slnx` by `generate-slnx.ps1`,
so there is nothing static to edit. Nothing under `.github/workflows`,
and no `.ps1`/`.yml` in `.azurepipelines`, names these projects.

The migration analyzer's shim tree mirrors the source project layout by
its own documented convention, so `Gds.Client.Common/` becomes
`Gds.Client/` and its readme table follows.

## Deliberately not changed

Prose naming past work: `plans/23` still records that its deferral
happened during the "GDS Client.Common modernization", which is the name
that effort had. File *paths* in the same document were updated.

## Validation

`net10.0`:

- `UA.slnx`: 0 errors. 14 warnings, all pre-existing
CA1861/CA1850/CA1307 in `XRegistry`, `Types.Tests` and `WotCon.Tests` —
none in anything this PR touches.
- Assemblies build as `Opc.Ua.Gds.dll`, `Opc.Ua.Gds.Client.dll`,
`Opc.Ua.Gds.Server.dll`, with no `*.Common.dll` produced anywhere.
- `Opc.Ua.Gds.Tests`: 1,116 passed, 0 failed, 54 skipped.
- `Opc.Ua.MigrationAnalyzer.Core.Tests`: 10 passed, 3 skipped.
- A repo-wide grep for the old names returns only the intentional old →
new mappings in the migration docs.

## Checklist

- [x] Existing tests pass
- [x] No new tests — this is a rename with no behaviour change
- [x] Consumer-visible package rename documented in the migration guide

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
2026-09-07 20:08:54 +02:00
Marc Schier 5b9bcedc62 Add OPC UA ISA-95 Common Model and Job Control support (#4089)
# Description

Adds first-class support for the OPC UA ISA-95 companion specifications:

- OPC-10030 Common Model.
- OPC-10031-4 Job Control V1.
- OPC-10031-4 Job Control V2.

The change introduces `Opc.Ua.ISA95`, `Opc.Ua.ISA95.Client`, and
`Opc.Ua.ISA95.Server`, together with a NativeAOT-compatible minimal
server sample and comprehensive tests.

## Highlights

- Compiles the three canonical NodeSets and identifier mapping CSV files
through the source generator.
- Adds NodeSet identifier mapping validation, typed method
arguments/results, companion event records, and generated method
argument nodes.
- Provides typed Common Model builders with normative relationships and
an injectable `GeoSpatialLocationType` provider.
- Implements lossless V1/V2 Job Control providers over a shared
version-neutral store, including state paths, status events, comments,
response queries, catalog changes, and cohesive DI registration.
- Adds subtype-aware typed clients, standalone encodeable registration,
`IIsa95ClientFactory`, and V2 status-event streaming.
- Adds the `MinimalIsa95Server` sample, package documentation, and an
ISA-95 guide with conformance and HA guidance.
- Tracks typed RSL/GPOS integration as follow-up issue #4095.

## Validation

- `Opc.Ua.ISA95.Tests`: 137/137 passed on `net10.0` and `net48`.
- Source-generator suites: 81/81 model, 90/90 stack, 3718 passed / 8
skipped core.
- `EventRecordFieldReaders` feedback tests: 6/6 on `net10.0` and
`net48`.
- `EventRecordFieldReaders` changed-method patch coverage: 100% line
coverage (36/36) and 100% branch coverage (22/22).
- `Opc.Ua.ISA95`, `.Client`, and `.Server` build across all configured
target frameworks without new diagnostics.
- Cobertura line coverage: 88.08% model, 90.04% client, 91.19% server.
- NativeAOT sample publishes and passes a live typed V2
Store/status-event validation.
- The ISA-95 NativeAOT smoke test passes in the source-generated native
test executable.
- OPC UA KB NodeSet validation reports zero errors for Common, V1, and
V2.

The broader NativeAOT executable still reports unrelated existing GDS
and complex-type test failures; the ISA-95-specific native test passes.

## Node state generator fixes

Generating method arguments inline for imported nodesets regressed every
model that already emits `InputArguments` / `OutputArguments` child node
factories:

- The inline block ran before the child factories.
`MethodState.CreateOrReplaceInputArguments` keeps the property that was
assigned first, so the child carrying the well-known NodeId was dropped
and a NodeId-less placeholder remained in the address space. This broke
browsing (reference descriptions without a target) and made method
arguments unreachable at their well-known NodeIds.
- The `MethodDeclarationId` fallback keyed on an autogenerated display
name redirected every declared method of the standard address space to
its own NodeId, which made method calls fail.

The inline block is now emitted after the child factories and reduced to
a value assignment when a child factory exists. The value is only
re-emitted when an argument references a namespace other than OPC UA,
because the value of an imported property is the verbatim XML of the
source nodeset whose namespace indices do not survive the import. The
`MethodDeclarationId` fallback is restricted to declarations without a
resolvable NodeId. The generated OPC UA address space is byte-identical
to `master` again.

## Related Issues

- Fixes #4088
- Follow-up: #4095
2026-07-28 12:01:57 +02:00
Marc Schier 7cec2d614d Move repository files to the consolidated layout 2026-07-16 06:25:54 +02:00