faster ⚡

[fair dawn]

Ran hyperfine to test dagger based go build vs go build locally. 10 iterations, 1 prerun

• dagger: mean 7.9 seconds
• go: 2.0

I've added the prior recommended go mod + build cache and that's the final result. I'm suspecting copying the assets into the container from my repo is the problem. Still before I flip all build actions to dagger I need to tackle this more

[sturdy saffron]

Have you been able to confirm that it's using the mod + build cache? I've had trouble putting them at the right path before.

Could also help to use include/exclude paths to limit what assets get copied

[fair dawn]

did dockerignore support get added?

[frank fern]

No, but @frozen talon built a helper: https://github.com/marcosnils/dagger-libs/blob/faeceba72d9951f8ce5e6bac0039a3a0496a4eff/ddocker/utils.go

[frozen talon]

@fair dawn @sturdy saffron I guess hyperfine also takes into account the session bootstrap process which is the one that Alex improved recently. I'd assume it's worth making testing with tip now?

[fair dawn]

https://gist.github.com/sheldonhull/84f45fc521274f144c01f3a02d494298

Here's an example of the code I used. Maybe I got it wrong.

To get the latest "tip" I should be updating my Go module dagger.io dependency, not dagger the binary correct?

[sturdy saffron]

hm you'd need the new dagger binary, since it's an improvement to code run by dagger session

the mounts LGTM, but it could help to use include/exclude here: https://gist.github.com/sheldonhull/84f45fc521274f144c01f3a02d494298#file-magefile-go-L49

something like **/*.go **/go.sum **/go.mod

[frozen talon]

yes, that's correct since each time you're building you're exporting something here https://gist.github.com/sheldonhull/84f45fc521274f144c01f3a02d494298#file-magefile-go-L78, that'll be copied back to the next build since you're selecting the entire CWD.

[fair dawn]

Include can support wildcard for all go files? I don't have nested under a simple root directory so it's harder to add a big list and maintain this