ci(vcpkg): Limit setup to vcpkg presets

[bobtista]

build-toolchain.yml runs run-vcpkg for every matrix preset even though only the win32-vcpkg* presets use the vcpkg toolchain. With executable caching disabled, every job bootstraps vcpkg independently, so transient GitHub release download failures make unrelated VC6 and ordinary Win32 builds fail.

Now vcpkg setup and binary-cache steps run only for presets containing vcpkg, the vcpkg executable and repository data are cached, and run-vcpkg is updated to the Node 24-based v11.6 release. The active VC6 and ordinary Win32 matrices no longer depend on vcpkg setup.

Todo:

• Parse the workflow as YAML
• Run git diff --check
• Confirm GenCI passes
• Replicate to Generals — N/A, shared CI workflow

[qodo-free-for-open-source-projects]

PR Summary by Qodo

CI: Run vcpkg setup only for vcpkg presets and enable caching

⚙️ Configuration changes ✨ Enhancement 🕐 10-20 Minutes

AI Description

• Gate all vcpkg setup/cache steps behind presets containing "vcpkg" to avoid unnecessary
 bootstraps.
• Enable caching for the vcpkg repo/executable and binary cache to reduce transient download
 failures.
• Bump lukka/run-vcpkg to v11.6 and stop using doNotCache for vcpkg presets.

Diagram

graph TD
  A["Matrix preset"] --> B{"Preset contains vcpkg?"}
  B -->|"No"| G["Build preset"]
  B -->|"Yes"| C["Compute cache key"] --> D["Restore bin cache"] --> E["run-vcpkg v11.6"] --> F["Configure cache dir"] --> G
  G --> H["Save bin cache (win32-vcpkg-debug only)"]

Loading

High-Level Assessment

The following are alternative approaches to this PR:

1. Add an explicit matrix flag (uses_vcpkg)

• ➕ Avoids relying on substring matching ("contains(preset, 'vcpkg')")
• ➕ More robust if preset naming conventions change
• ➕ Makes intent clearer in the workflow/matrix definition
• ➖ Requires updating matrix generation/definition to carry the flag
• ➖ Slightly more verbose workflow wiring

2. Split vcpkg into a dedicated job reused by vcpkg presets

• ➕ Hard separation: non-vcpkg builds cannot be impacted by vcpkg setup
• ➕ Can centralize caching and vcpkg bootstrap logic
• ➕ Potentially parallelizes/isolates failures better
• ➖ More workflow complexity (job outputs/artifacts, dependencies)
• ➖ May increase runtime due to job orchestration overhead

Recommendation: Current approach is a good pragmatic fix: gating vcpkg work to vcpkg presets and enabling caching directly addresses flaky bootstrap failures and reduces wasted work. If preset naming is expected to evolve, consider the explicit matrix flag approach for long-term robustness; otherwise, the substring gate is sufficient and low-risk.

Files changed (1) +7 / -6

Other (1) +7 / -6

build-toolchain.ymlGate vcpkg steps to vcpkg presets and enable caching; bump run-vcpkg +7/-6

Gate vcpkg steps to vcpkg presets and enable caching; bump run-vcpkg

• Updates vcpkg cache-key computation, cache restore, vcpkg setup, and cache configuration steps to run only when the preset name contains "vcpkg". Upgrades lukka/run-vcpkg to v11.6 and enables its caching (doNotCache: false). Adjusts the cache save condition to remain single-writer while still restricted to vcpkg presets.

.github/workflows/build-toolchain.yml

[qodo-free-for-open-source-projects]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0)

Great, no issues found!

Qodo reviewed your code and found no material issues that require review

Tip of the day

💡 Did you know, you can type 'qodo, fix this' on a finding and the fix lands right on your PR

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗

[xezon]

ok lets go