[robherley]

When we create cache entries, it's normal behavior for multiple jobs to "race" to create the same entry. In the cache v1 implementations, we suppressed these errors with just a normal log statement however in v2 we create an annotation with warning that is confusing folks.

Here's the v1 implementation:

toolkit/packages/cache/src/cache.ts

Lines 402 to 425 in 662b9d9

	const reserveCacheResponse = await cacheHttpClient.reserveCache(
	key,
	paths,
	{
	compressionMethod,
	enableCrossOsArchive,
	cacheSize: archiveFileSize
	}
	)
	if (reserveCacheResponse?.result?.cacheId) {
	cacheId = reserveCacheResponse?.result?.cacheId
	} else if (reserveCacheResponse?.statusCode === 400) {
	throw new Error(
	reserveCacheResponse?.error?.message ??
	`Cache size of ~${Math.round(
	archiveFileSize / (1024 * 1024)
	)} MB (${archiveFileSize} B) is over the data cap limit, not saving cache.`
	)
	} else {
	throw new ReserveCacheError(
	`Unable to reserve cache with key ${key}, another job may be creating this cache. More details: ${reserveCacheResponse?.error?.message}`
	)
	}

In v2, our client throws an error for a failed request, so we need to wrap that in the ReserveCacheError. That ensures we only emit an info statement, not a warning annotation:

toolkit/packages/cache/src/cache.ts

Lines 564 to 565 in 662b9d9

	} else if (typedError.name === ReserveCacheError.name) {
	core.info(`Failed to save: ${typedError.message}`)