Improve cache service availability determination and implement conditional error logging

[Copilot]

This PR improves how the cache service availability is determined in @actions/toolkit and implements conditional error logging for cache failures based on HTTP status codes.

Changes Made

1. Enhanced isFeatureAvailable() function

The isFeatureAvailable() function now properly leverages the ACTIONS_CACHE_SERVICE_V2 feature flag to determine cache service availability:

• For cache service v2: Requires ACTIONS_RESULTS_URL to be set
• For cache service v1: Requires ACTIONS_CACHE_URL to be set
• For GHES: Always uses v1 logic regardless of the v2 flag setting

Before:

export function isFeatureAvailable(): boolean {
  return !!process.env['ACTIONS_CACHE_URL']
}

After:

export function isFeatureAvailable(): boolean {
  const cacheServiceVersion = getCacheServiceVersion()
  
  switch (cacheServiceVersion) {
    case 'v2':
      return !!process.env['ACTIONS_RESULTS_URL']
    case 'v1':
    default:
      return !!process.env['ACTIONS_CACHE_URL']
  }
}

2. Conditional error logging for cache failures

Updated cache failure logging to differentiate between server errors and other failures:

• Server errors (5xx status codes): Logged as errors using core.error()
• All other cache failures: Logged as warnings using core.warning()

Applied to all cache functions:

• restoreCacheV1() function
• restoreCacheV2() function
• saveCacheV1() function
• saveCacheV2() function

Implementation:

// Log server errors (5xx) as errors, all other errors as warnings
if (
  typedError instanceof HttpClientError &&
  typeof typedError.statusCode === 'number' &&
  typedError.statusCode >= 500
) {
  core.error(`Failed to restore: ${(error as Error).message}`)
} else {
  core.warning(`Failed to restore: ${(error as Error).message}`)
}

3. Comprehensive test coverage

Added extensive test coverage including:

• V1 service with ACTIONS_CACHE_URL requirements
• V2 service with ACTIONS_RESULTS_URL requirements
• GHES behavior with v2 flag enabled
• Server error (5xx) vs other error logging differentiation
• Edge cases with no URLs set
• Updated existing tests to expect appropriate log levels

Backward Compatibility

This change maintains full backward compatibility:

• Existing workflows using ACTIONS_CACHE_URL continue to work unchanged
• V1 cache service behavior is preserved
• GHES installations are unaffected
• Only server errors (5xx) are elevated to error level, preserving existing warning behavior for most cache failures

Fixes #2099.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[Link-]

Fix the implementation as per the comments provided

[Copilot]

Fix the implementation as per the comments provided

Implemented as requested in commit d48d6b6. Removed the logCacheError function and added inline 5xx server error detection in all four cache functions.

[Link-]

Environment	Screenshot
github.com
GHES 3.16

[Link-]

@copilot Update the PR description to match the changes introduced

[Copilot]

@copilot Update the PR description to match the changes introduced

Updated the PR description to accurately reflect the actual implementation. The description now correctly describes:

1. The exact logic for isFeatureAvailable() - v1 only checks ACTIONS_CACHE_URL, v2 checks ACTIONS_RESULTS_URL
2. Conditional error logging - only server errors (5xx) are logged as errors, other failures remain warnings
3. The comprehensive test coverage for all scenarios

The updated description matches the current implementation rather than the original planned changes.