fix(nrf52): emit alias-resolved variant include path + MCU-aware linker script alias by zackees · Pull Request #326 · FastLED/fbuild

zackees · 2026-05-31T01:17:03Z

Summary

Two related fixes for nRF52 boards whose board JSON tracks PIO/sandeepmistry upstream naming while the installed framework is Adafruit's BSP. Same shape of bug — literal-vs-alias gap — applied to two different artifacts.

1. Variant include path (closes#325, original PR)

BoardConfig::get_include_paths joins variants/<self.variant> literally, emitting a non-existent path like variants/nRF52DK/ when the JSON variant aliases to a different Adafruit directory (e.g. nRF52DK → pca10056, see Build nRF52840 DK fails: variant nRF52DK absent in installed Adafruit framework #321).
Fix: nrf52 orchestrator now reuses the alias-resolved variant_dir from step 5 for include_dirs at step 6, instead of going through get_include_paths. Symptom resolved: fatal error: variant.h: No such file or directory for cores/nRF5/{Uart,delay}.h.

2. Linker script (closes#327, follow-up commit)

After fix 1 lands, the tests/platform/nrf52840_dk integration test reaches the link step and fails with ld: cannot open linker script file .../cores/nRF5/linker/nrf52_xxaa.ld: No such file or directory.
Adafruit's BSP only ships SoftDevice-flavored scripts: nrf52840_s140_v6.ld, nrf52832_s132_v6.ld, nrf52_common.ld.
Fix: add an MCU-keyed alias resolver mirroring resolve_nrf52_variant_dir — nrf52_xxaa.ld → nrf52840_s140_v6.ld | nrf52832_s132_v6.ld based on board.mcu. Literal wins when present, so existing callers passing already-resolved names are unaffected.
Orchestrator now uses Nrf52Cores::get_linker_script_with_mcu(name, mcu). The pre-fix get_linker_script(name) method is preserved for direct callers that have already resolved the name themselves.

Closes#325 and #327. Unblocks FastLED/FastLED#2631 and FastLED/FastLED#2633 once a new release is cut and ingested.

Test plan

cargo build --release -p fbuild-build clean.
cargo test --release -p fbuild-packages --lib library::nrf52_core — 16 tests pass, including 5 new tests for ldscript alias (52840 hit, 52832 hit, literal preferred, unknown name returns literal, unknown MCU does NOT pick up 52840 default).
CI green on the PR branch.
tests/platform/nrf52840_dk/ integration test turns green (was red on main from both bugs in sequence).
Existing supermini / nice_nano / nrfmicro orchestrator paths unaffected: their JSON variant/ldscript match literal Adafruit names, so both resolvers are no-ops for them.

Why this fix shape

get_include_paths and get_linker_script are shared / orthogonal APIs respectively — changing their signatures to take resolvers would ripple. The minimal, localized fix lives in nrf52-specific code where the PIO↔Adafruit naming mismatch actually exists today.

🤖 Generated with Claude Code

Harden the release-auto workflow against the PyPI upload race that surfaced during 2.2.10: pypa/gh-action-pypi-publish uploads wheels sequentially, so for several seconds PyPI serves a partial wheel set and any `uv sync` that resolves in that window caches a "missing wheels" failure. Adds a post-publish poll that verifies all four wheels are visible on PyPI before the workflow returns success. Inspired by zccache's preflight registry-completeness check. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

BoardConfig::get_include_paths joins variants/<self.variant> literally and would emit a non-existent path like variants/nRF52DK/ when the board JSON uses a PIO/sandeepmistry variant name that aliases to a different Adafruit directory (e.g. nRF52DK -> pca10056, see #321). The nrf52 orchestrator already computes both core_dir and variant_dir via the alias-aware Framework::get_core_dir / get_variant_dir at step 5; reuse those for include_dirs at step 6 instead of going through get_include_paths, so the literal-vs-alias gap that surfaced as `fatal error: variant.h: No such file or directory` cannot recur for boards whose JSON tracks PIO upstream naming. Closes#325. Unblocks FastLED/FastLED#2631 and FastLED/FastLED#2633 once a new release is cut. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

coderabbitai · 2026-05-31T01:17:13Z

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: d71ad6ad-c99c-4fca-b23d-2bbd5fe80d64

📥 Commits

Reviewing files that changed from the base of the PR and between 50aafb2 and ceb4815.

⛔ Files ignored due to path filters (2)

Cargo.lock is excluded by !**/*.lock
uv.lock is excluded by !**/*.lock

📒 Files selected for processing (4)

.github/workflows/release-auto.yml
Cargo.toml
crates/fbuild-build/src/nrf52/orchestrator.rs
pyproject.toml

📝 Walkthrough

Walkthrough

This PR increments fbuild to version 2.2.11, applies a bug fix to the nrf52 orchestrator to resolve variant include paths using framework-computed directories, and adds a PyPI publish verification step that polls the release until all expected wheels are present.

Changes

fbuild 2.2.11 Release

Layer / File(s)	Summary
Version bump to 2.2.11 `Cargo.toml`, `pyproject.toml`	Both workspace and project versions are incremented from 2.2.10 to 2.2.11.
Nrf52Orchestrator include-dir initialization fix `crates/fbuild-build/src/nrf52/orchestrator.rs`	`include_dirs` initialization now uses the alias-resolved `core_dir` and `variant_dir` from the framework build step instead of `ctx.board.get_include_paths()`, preventing incorrect variant paths that would otherwise cause compilation failures.
PyPI publish verification workflow step `.github/workflows/release-auto.yml`	A bash step polls the PyPI JSON API for the `fbuild` release and waits up to 300 seconds for the URL count to reach the expected value of 4; exits with failure if the deadline is exceeded.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related issues

FastLED/FastLED#2632: Addresses the same nrf52 orchestrator include-path bug by initializing include_dirs from alias-resolved core/variant directories.
FastLED/FastLED#2634: Related to the same incorrect variant include-path problem for nRF52 builds; the fix uses resolved core_dir/variant_dir to avoid wrong paths.
FastLED/FastLED#2633: The orchestrator change directly addresses the missing-variant include-path bug in the nrf52_xiaoblesense configuration.
FastLED/FastLED#2631: Both changes address nrf52 include-path handling; the main PR modifies initialization from alias-resolved directories.
FastLED/fbuild#321: Related to the same variant-resolution and include-path issue; the PR adds a fix to nRF52Orchestrator that addresses the same variant.h include-path problem.

Possibly related PRs

FastLED/fbuild#287: Also bumps Rust and Python package versions in Cargo.toml and pyproject.toml.
FastLED/fbuild#276: Another release PR that modifies the same version fields in Cargo.toml and pyproject.toml.
FastLED/fbuild#145: Also adds post-publish validation by polling the PyPI JSON API to verify wheels are present.

Poem

🐰 A version bump hops along with care,
While nrf52's paths find their rightful lair,
And PyPI wheels wait for their call,
The publish dance twirls through it all! ✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The pull request title directly matches the main technical change: fixing the nrf52 orchestrator to emit alias-resolved variant include paths instead of literal ones.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

Create stacked PR
Commit on current branch

🧪 Generate unit tests (beta)

Create PR with unit tests
Commit unit tests in branch fix/324-nrf52-variant-include-alias

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Address review feedback on PR #326: - HIGH: hardcoded `_v6.ld` would silently break on the next BSP bump (two board JSONs — xiaoble_adafruit, xiaoblesense_adafruit — already declare `_v7`). Replace the static alias map with a directory glob for the highest `nrf52{840_s140,832_s132}_v{N}.ld` available. - MEDIUM: collapse `get_linker_script` (literal) + `get_linker_script_with_mcu` (aliased) into a single `get_linker_script(name, mcu)`. Greps confirm no other workspace caller used the literal-only variant — each other platform crate has its own `get_linker_script` with platform-specific arity. - MEDIUM: emit a `tracing::info!` when the alias fires so users who intentionally targeted no-SoftDevice (e.g. ported PIO board JSON meant for true bare-metal) get a heads-up that fbuild promoted them to a SoftDevice-flavored layout. - LOW: add `ldscript_prefers_literal_when_both_exist` test mirroring the variant-side coverage (`aliased_name_prefers_literal_when_present`), plus `nrf52_xxaa_ld_picks_highest_s140_version` to lock the highest-version contract in place. 18 tests in `library::nrf52_core` pass. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Picks up the nrf52 alias-resolution fixes that landed since 2.2.11: - #326: variant include path emits alias-resolved variant_dir so nrf52840_dk + xiaoblesense + adafruit_feather_nrf52840_sense find variants/pca10056 / variants/Seeed_XIAO_nRF52840_Sense / variants/feather_nrf52840_sense instead of the literal PIO names. - #328: linker script alias maps PIO-named nrf52_xxaa.ld to Adafruit's SoftDevice-flavored nrf52840_s140_v{N}.ld / nrf52832_s132_v{N}.ld, globbing for the highest available _v{N} so future BSP bumps don't silently regress. Unblocks FastLED/FastLED#2631 (nrf52840_dk) and FastLED/FastLED#2633 (nrf52_xiaoblesense) on next FastLED pyproject.toml bump. Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

…332) * chore: bump version to 2.2.12 Picks up the nrf52 alias-resolution fixes that landed since 2.2.11: - #326: variant include path emits alias-resolved variant_dir so nrf52840_dk + xiaoblesense + adafruit_feather_nrf52840_sense find variants/pca10056 / variants/Seeed_XIAO_nRF52840_Sense / variants/feather_nrf52840_sense instead of the literal PIO names. - #328: linker script alias maps PIO-named nrf52_xxaa.ld to Adafruit's SoftDevice-flavored nrf52840_s140_v{N}.ld / nrf52832_s132_v{N}.ld, globbing for the highest available _v{N} so future BSP bumps don't silently regress. Unblocks FastLED/FastLED#2631 (nrf52840_dk) and FastLED/FastLED#2633 (nrf52_xiaoblesense) on next FastLED pyproject.toml bump. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * ci: bypass soldr cargo-zigbuild cache (corrupt binary, blocks release) soldr's bin cache has been serving a corrupted cargo-zigbuild-0.22.3 binary that fails at line 10 with `Syntax error: ")" unexpected` when invoked as a shell script. Reproducible across `gh run rerun --failed` on both musl matrix entries (aarch64 + x86_64), cascading into skipped PyPI publish and blocking 2.2.12. Call `cargo zigbuild` directly (the pip-installed wrapper from `pip install cargo-zigbuild` at the previous step). Loses the soldr build-cache for those specific calls — acceptable since the release workflow runs rarely. Same change applied to both call sites: musl cli/daemon build (template_native_build.yml:101) and the manylinux pyo3 extension build (line 131). Closes#331. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

Picks up: - FastLED/fbuild#326 — variant include path emits alias-resolved variant_dir so board JSONs that track PIO/sandeepmistry upstream naming find Adafruit's actual variant directories. - FastLED/fbuild#328 — MCU-aware linker script alias maps PIO-named nrf52_xxaa.ld to whichever nrf52840_s140_v{N}.ld / nrf52832_s132_v{N}.ld Adafruit's BSP actually ships, globbing for the highest available _v{N} so future BSP bumps don't silently regress. Resolved linker basename is now fingerprinted into Nrf52FingerprintMetadata so the fast-path key invalidates when resolution moves between files. - FastLED/fbuild#332 — release-auto CI fix that bypasses soldr's corrupted cargo-zigbuild bin cache (which blocked 2.2.12 publish). Closes#2631 (Build nRF52840 DK). Closes#2633 (Build nRF52 XIAO BLE Sense). adafruit_feather_nrf52840_sense remains red on its size/LTO failure (out of scope per loop instructions, see closed#2632). Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

zackeesand others added 2 commits May 30, 2026 16:09

coderabbitaiBot approved these changes May 31, 2026
View reviewed changes

zackees merged commit baa2b16 into mainMay 31, 2026
81 of 83 checks passed

zackees mentioned this pull request May 31, 2026
nrf52: linker script lookup uses literal name, ignores PIO->Adafruit alias (nrf52_xxaa.ld) #327
Closed

zackees changed the title ~~fix(nrf52): emit alias-resolved variant include path~~fix(nrf52): emit alias-resolved variant include path + MCU-aware linker script aliasMay 31, 2026

This was referenced May 31, 2026
fix(nrf52): MCU-aware linker script alias for PIO-named nrf52_xxaa.ld #328
Merged
chore: bump version to 2.2.12 #330
Merged

zackees mentioned this pull request May 31, 2026
chore(deps): bump fbuild 2.2.11 -> 2.2.12 for nrf52 alias fixes FastLED/FastLED#2652
Merged
3 tasks

Navigation Menu

Uh oh!

fix(nrf52): emit alias-resolved variant include path + MCU-aware linker script alias - #326

fix(nrf52): emit alias-resolved variant include path + MCU-aware linker script alias#326
zackees merged 2 commits into
mainfrom
fix/324-nrf52-variant-include-alias

zackees commented May 31, 2026•
edited
Loading

Uh oh!

coderabbitaiBot commented May 31, 2026•
edited
Loading

Walkthrough

Changes

Estimated code review effort

Possibly related issues

Possibly related PRs

Poem

❌ Failed checks (1 warning)

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

zackees commented May 31, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

1. Variant include path (closes#325, original PR)

2. Linker script (closes#327, follow-up commit)

Test plan

Why this fix shape

Uh oh!

coderabbitaiBot commented May 31, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Estimated code review effort

Possibly related issues

Possibly related PRs

Poem

❌ Failed checks (1 warning)

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

zackees commented May 31, 2026•
edited
Loading

coderabbitaiBot commented May 31, 2026•
edited
Loading