Repository navigation
Action takes 5-6 seconds to download #2542
Description
Activity
Hi @jsoref,
Thanks for your suggestion. I will forward it to the team responsible for the Action and provide an update when I receive more information.
Reacted by Josh SorefNow that #3054 is merged, download times should be significantly improved. Thanks for the effort you've put into this already. We'll close this issue and please let us know if download times continue to be a problem.
Confirmed. 14s->1s
before (14s)
Fri, 29 Aug 2025 17:35:24 GMT Download action repository 'github/codeql-action@48dd624a81acd5d5f2e94c2b1e54102c6b5bd642' (SHA:48dd624a81acd5d5f2e94c2b1e54102c6b5bd642) Fri, 29 Aug 2025 17:35:38 GMT Complete job name: timeafter (1s)
Fri, 29 Aug 2025 17:35:49 GMT Download action repository 'github/codeql-action@02ab253bd299d261d00cdf8a9bca38fea2697d50' (SHA:02ab253bd299d261d00cdf8a9bca38fea2697d50) Fri, 29 Aug 2025 17:35:50 GMT Complete job name: timeNice!
Fwiw, it looks like you now have 45mb of content (42mb is the bundled actions, and 3mb is everything else) which is a huge improvement over what it was before.
Personally, I'd kinda want repositories or tags that had only a single action since that's the difference between 2-5M and 45M -- and I'm pretty sure it'd be doable if you were willing to create things where you basically deleted everything except the top level readmes, the relevant js file and the relevant action, committed, and tagged that as a single tag (either
${sha}-${action}or${action}-${sha}).But that's just a bit of cream and I can understand not wanting to do it...
I can see how that would be useful for
upload-sarifusers. It would be less useful for codeql code scanning users since they are the vast majority of codeql action users. I suspect they would see much less of an improvement if we split things up as you suggest.
It takes 5-6 seconds for workflows to download this repository as an action. I believe it's because the node modules directory is 300+mb. It includes at least two copies of typescript.
@aibaars suggests: