This message was deleted.

If it’s not ok to bring this up here, let me know. I figured it be ok given the announcement.

I don’t believe we want the

pulumi

command-line client to do too much “magic” behind the scenes with regard to running

npm install

,

yarn install

, etc. Since there are plenty of corner cases we’d probably have trouble getting right. (e.g. if you are using a custom package manager.) So I think we’d want to leave “gathering runtime dependencies” to whatever script or CI/CD environment you are using to execute

pulumi

. (So that then you can confirm you have the right setup.) But what I think we can/should/will do is make sure that we do have a good escape hatch so you can tie custom steps into the scripts or containers that we publish. For example, the

entrypoint.sh

for the

pulumi/pulumi

Docker container (like in your PR). Or supporting a generic

if [ -f pulumi-pre-update.sh ]

or something. Make sense? This is a problem we want to make sure gets solved, but it probably makes more sense to solve it upstream.

It already runs

npm install

though (hardcoded).

Hi @cool-egg-852! We do have an open issue to fix this for the policy packs! If you any input/suggestions, feel free to DM me and we can discuss more.

Sounds good, thanks.