It dramatically reduces the amount of time it takes to re-run image building if the set of packages used hasn't changed. It's hard to get an objective measurement (because we do get some caching from being careful what we add), but:
On my machine, for a rebuild after a change in 'pkg/api', it reduces make docker-build time from 249s -> 22s.
And, it also slightly improves the time it takes to run a clean build. After docker system prune -af, it reduces make docker-build on my machine from 345s -> 126s.
I'm not super enthusiastic about this particular arrangement, because we need to make sure that all the dockerfiles exactly match at the top, but maybe it's ok for now?
Also, this change affects our image building for other platforms, I think? It was inconsistent between neonvm components and the others anyways though, so I'm not entirely sure it matters.
sharnoff
commented
2 weeks ago
It dramatically reduces the amount of time it takes to re-run image building if the set of packages used hasn't changed. It's hard to get an objective measurement (because we do get some caching from being careful what we add), but:
On my machine, for a rebuild after a change in 'pkg/api', it reduces
make docker-buildtime from 249s -> 22s.And, it also slightly improves the time it takes to run a clean build. After
docker system prune -af, it reducesmake docker-buildon my machine from 345s -> 126s.I'm not super enthusiastic about this particular arrangement, because we need to make sure that all the dockerfiles exactly match at the top, but maybe it's ok for now?
Also, this change affects our image building for other platforms, I think? It was inconsistent between neonvm components and the others anyways though, so I'm not entirely sure it matters.