Repository navigation
fix(ci): provision cold Linux capacity through the Android SDK manager - #1892
Merged
Merged
Conversation
|
Important Review skippedAuto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the
|
This was referenced Nov 4, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fresh Linux builders for the 2.5.51 release started with only 11.91–12.48 GiB available. Legitimate cross SDK downloads and Rust outputs then hit the existing 5 GiB capacity guards: ESP32-P4 acceptance had 4.14 GB free, GNU glibc-floor cross-linking 1.86 GiB, macOS ARM sidecar repair 3.57 GiB, and Linux ARM musl extension preparation 2.37 GiB. The target cache was cold, so deleting or disabling restored Rust caches would not solve the provisioning gap.
Before those builds, uninstall only the unused Android NDK packages supplied by the GitHub-hosted Linux image. A shared CI helper reads installed SDK package metadata, validates the exact NDK IDs, and delegates removal to
sdkmanagerwith the explicit SDK root. It records workspace free bytes before and after. Its entrypoint refuses local, self-hosted, and non-Linux execution. The reusable native template also skips the operation on other runners.The helper is used by four existing workflow paths: acceptance, glibc-floor, the all-OS Dylint Linux builder, and Linux native release builds. The initial PR sweep exposed the same capacity guard in Dylint at 3.99 GiB while preparing its Windows cross-check, so that actual consumer is included too. Other Android packages, Rust/fbuild caches, and all disk guards remain unchanged. No workflow files or production Rust behavior are added.
Validation: five focused tests cover installed metadata selection, no-op behavior, invalid metadata, hosted-only guards, and actual subprocess arguments through a temporary SDK-manager recorder. Ruff, Python complexity, workflow concurrency checks, and independent review passed. The PR sweep measured about 6.49 GiB reclaimed from three installed NDKs; cold P4 acceptance and both GNU glibc-floor architectures passed. Both final-head Dylint jobs passed too: one started with 13,880,700,928 free bytes, reclaimed 6,968,242,176 bytes, and completed its cross-target checks with the 5 GiB guard enabled (5.87 GiB remained at the later check). The final head must complete the rest of its full sweep before merge.
ci-fullwas applied at creation. The subsequent exact-candidate release must also prove the native release cross-build profiles.The failing exact-candidate release is https://github.com/FastLED/fbuild/actions/runs/41808037950. This follows #1887 and #1888 and unblocks the 2.5.51 consumer cascade for FastLED/FastLED#5015.