Repository navigation
allow skipping validation #253
Description
ci runs in chisel releases need to install slices multiple times. every install runs on the same repo and is forced to run validation. this currently takes about 60% of the CI time. it woudl be an incredibly useful feature to allow us to skip validation in certain contexts. i've been running chisel quite successfully only for local dev-time testing with the following:
diff --git i/internal/setup/setup.go w/internal/setup/setup.go
index 4502687..1909f8e 100644
--- i/internal/setup/setup.go
+++ w/internal/setup/setup.go
@@ -187,6 +187,12 @@ func ReadRelease(dir string) (*Release, error) {
return nil, err
}
+ // skip validation for cached releases is CHISEL_HACKS flag is set
+ if os.Getenv("CHISEL_HACKS") != "" {
+ logf("Skipping release validation due to CHISEL_HACKS")
+ return release, nil
+ }
+
err = release.validate()
if err != nil {
return nil, err
(i don't propose this exact code to be merged)
Reacted by Brennan Kinney
Activity
letFunny commentedon Nov 11, 2025
Related to #254. In Chisel we never want to skip validation of the release, that will create more problems down the line. Think about working on a local version of Chisel cut and turning it off, then turning it back on only to see validation errors and think instead is better to just leave it off. It could happen just by copying code with the wrong flags from the CI. We would get many error reports and we would need to differentiate between the two leading to more frustrated users.
I do agree that it takes a long time at the moment, especially when running it on the CI. It is also true that validation is the single most important operation that we perform. While I do not think it is a good idea to disable it, I do think we could cache the result or speed it up.
upils commentedon Oct 22, 2026
With recent optimizations the validation step is now taking a negligible time (at least with the current set of packages). I am closing this issue for now, let's reopen it or open another if we ever need to revisit the idea of caching the validation.
lczyk commentedon Oct 23, 2026
linking this too as a breadcumb: