Open jamesward opened 4 years ago
I think the point of that currently is to separate build and push and render as separate steps so that users don’t keep staring at a static screen for a long time.
Also for a beginner, it highlights the fact that the image is built and is pushed to somewhere. That’s my current rationale, also some code reuse around pushing images.
With Jib right now we don't do the manual push because Jib does it for us. (I had just noticed that in the code.)
For timing, here are the results from my local machine on a Node app:
$ time pack build --publish gcr.io/crb-test/nodebars
real 0m27.635s
$ time pack build gcr.io/crb-test/nodebars
real 0m20.684s
So it does take 7 seconds longer to do it with --publish
. Which is a while on the when it doesn't seem like anything is happening. So maybe to do this we'd have to reconsider not displaying the build output. Or we can disable Jib's publish and make it work like the others for consistency.
I think it's cleaner to disable Jib's publish and use docker push
as well.
There was a time several years ago that pushing images from code wasn't very future-proof and had issues with credential providers etc, but now they probably work fine.
Regardless, I'm ok both ways. My main motive was to show some progress in the CLI by keeping them seperate.
We should be able to do
--publish
with the pack build and then not push the image manually.