Fix configs seed self-copy in all-users scope

[WaynePatrickDalton]

Fix configs seed self-copy failure in all-users scope

Description

Hey @jmcouffin - I came across this when updating our pyRevit wix based installer. Looking it up online, seems a few more users over the years have had this issue - although, perhaps this 'bug' is intended feature of pyrevit. Let me know eitherway!

This pull request attemps to fix a known bug where configs seed can fail with a misleading file-lock style error when source and target resolve to the same config file in all-users scope. This PR was generated using GPT5.3 Codex and Fable5 via Github Copilot on VSC.

What changed:

• Normalizes source and target paths using Path.GetFullPath, trims trailing directory separators, and compares using OrdinalIgnoreCase.
• Adds same-path handling in SeedConfig:
• Skips copy when source equals target.
• Logs a debug message explaining seeding was skipped because both paths resolve to the same file.
• Returns success instead of throwing.
• Preserves lock intent for same-path behavior by setting read-only on the target when lockSeedConfig is true and the file exists.
• Preserves different-path behavior and existing exception wrapping semantics for real failures.
• Adds unit tests for:
• same-path seed success without throw,
• same-path seed with lock applying read-only,
• different-path seed still copying as before.

Background and prior reports:

• Discourse report(s):
  configs-seed-error
  cloning-pyrevit-and-our-extensions-for-deployment-with-powershell-doesnt-work
• GitHub issue: pyRevit issue #1266

---

Checklist

• Code follows the PEP 8 style guide. (Not applicable for this C# change)
• Code has been formatted with Black. (Not applicable for this C# change)
• Changes are tested and verified to work as expected.

Validation performed:

• dotnet build -c Debug
• dotnet vstest .\bin\Debug\net48\pyRevitLabs.UnitTests.dll --TestCaseFilter:"FullyQualifiedName~PyRevitConfigsTests"
• dotnet vstest .\bin\Debug\net48\pyRevitLabs.UnitTests.dll --TestCaseFilter:"FullyQualifiedName~PyRevitInstallScopeTests"

Result summary:

• Build succeeded with 0 warnings and 0 errors.
• SeedConfig tests passed: 3 passed, 0 failed.
• Install scope regression sanity tests passed: 6 passed, 0 failed.

---

Related Issues

• Resolves #1266
• Related community discussion: configs seed error

---

Additional Notes

Reference documentation:

• pyRevit Configuration
• pyRevit Seed Config

Implementation note:

• The fix targets the all-users self-copy case specifically (ProgramData to the same ProgramData file) and keeps the existing copy/failure behavior for true different-path operations.

[devloai]

PR Summary:

• Fixes a bug where SeedConfig fails with a misleading file-lock error when source and target resolve to the same config file in all-users scope (ProgramData → same ProgramData file).
• Normalizes source/target paths via Path.GetFullPath, compares with OrdinalIgnoreCase, and skips the copy when they're identical — logging a debug message and returning success instead of throwing.
• Preserves lock intent by applying read-only attributes on the target when lockSeedConfig is true.
• Extracts the lock logic into a shared ApplySeedConfigLock helper to avoid duplication.
• Adds unit tests covering same-path skip, same-path with lock, and different-path copy behavior.

Review Summary:

The fix is well-targeted and correctly addresses issue #1266. The path normalization and same-path short-circuit logic is sound, and the refactoring into ApplySeedConfigLock is clean. One medium-severity issue: NormalizePath (which calls Path.GetFullPath) is invoked outside the try block that wraps failures in PyRevitException, so malformed paths will propagate unwrapped System.ArgumentException instead of the expected PyRevitException — this breaks the PR's stated goal of preserving exception-wrapping semantics. The fix is to move the samePath computation inside the existing try. One low-severity observation: the same-path branch silently skips the lock when the config file doesn't exist, with no diagnostic log explaining that the lock wasn't applied. Tests are well-structured with proper setup/cleanup and environment variable isolation.

Suggestions

• Add a test case for SeedConfig when sourceFile does not exist in same-path mode to verify it returns without throwing. Apply
• Add a test verifying that SeedConfig wraps Path.GetFullPath failures in PyRevitException after moving NormalizePath inside the try block. Apply

[WaynePatrickDalton]

Pulling the PR branch fast-forwarded in upstream develop changes that reworked active-config resolution (GetActiveConfig now routes non-elevated processes to the per-user config), which broke the two same-path tests. I updated pyRevitLabs.UnitTests.PyRevitConfigs.cs to drive the same-path scenario via an admin-locked (read-only) machine config, which resolves deterministically regardless of elevation.

Validation:

Build: 0 Warning(s) 0 Error(s)
PyRevitConfigsTests: Passed! - Failed: 0, Passed: 3
PyRevitInstallScopeTests: Passed! - Failed: 0, Passed: 14, Skipped: 3 (skips are elevation-dependent tests, pre-existing)