Uh oh!
There was an error while loading. Please reload this page.
feat(symbols): map-derived synthesis for anonymous rodata pools (closes #425) - #427
Merged
Conversation
#425) `fbuild symbols` previously stopped at object-file granularity for `.flash.rodata` because `nm` does not enumerate the anonymous merged string pools that `-fdata-sections` puts in `.rodata.<mangled>.str1.<N>` sub-sections. Concretely, ~377 KB of the ESP32-S3 Blink master ELF's rodata fell back to "(archive, object, .flash.rodata) -> N bytes" rather than "this function's strings -> N bytes". The linker map carries the owner inline in the input-section name. This PR adds: - `extract_owner_from_section()` in `fbuild-core::symbol_analysis` that recognises the input-section shapes that carry an owner: `.text.<owner>`, `.literal.<owner>`, `.rodata.<owner>`, `.rodata.<owner>.str1.<N>`, `.rodata.<owner>.cst<N>`, `.data.<owner>`, `.bss.<owner>`, `.data.rel.ro.<owner>`, `.gnu.linkonce.{t,r,d,b}.<owner>`. - `collect_map_derived_owners()` that walks map ranges, skips any range whose address is covered by an nm-listed symbol (so we don't double-count), and returns the owner+size for each remaining range. - `build_fine_grained_map_with_synth()` (and a thin wrapper for the existing call site) that emits synthetic `FineGrainedSymbol` rows for those owners, tagged `source: "map-derived"`. - New `source: "nm" | "map-derived"` field on `FineGrainedSymbol` with a `serde(default)` so JSON written by earlier `fbuild symbols` still deserialises. - The CLI driver (`fbuild-build::symbol_analyzer::analyze_elf`) collects synthetic mangled owners and demangles them via c++filt in the same threaded batch as the nm names. Empirically on the same ESP32-S3 Blink master ELF that motivated #425: | Metric | Before | After | |---|---:|---:| | Sized symbols emitted | 3,696 | 4,591 (+895 map-derived) | | Flash bytes named at symbol level | 316,368 | 388,380 (+72 KB) | The single largest new attribution: 58 KB of merged rodata in `main.cpp.o` now correctly named as belonging to `fl::ClocklessIdf5<3, fl::TIMING_WS2812_800KHZ, GRB, 0, false, 280>::ClocklessIdf5`, which is the FL_WARN string pool pulled in by the NEOPIXEL chipset header. Per-archive rodata for FastLED audio/channels/gfx/stl/net detector classes also now resolves to function-level owners. Tests ----- 11 new unit tests in `fbuild-core::symbol_analysis::tests` covering all input-section shapes listed above, the no-double-count invariant, range/coverage overlap logic, and the synthetic-row pipeline end to end. Plus 2 existing tests in `fbuild-build::symbol_analyzer::tests` (toolchain prefix derivation). All 24 unit tests across both crates pass. Verified end-to-end on real ESP32-S3 ELFs. Closes#425.
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
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. Comment |
…s' into feat/425-rodata-symbol-attribution
…OC gate PR #427 surfaced the workspace's "Reject .rs files over 1000 LOC" gate because the new map-derived synthesis code + its 13 unit tests pushed crates/fbuild-core/src/symbol_analysis.rs to 1076 lines (limit: 1000). Promote `symbol_analysis` from a single file to a directory module: - crates/fbuild-core/src/symbol_analysis/mod.rs (695 LOC) - crates/fbuild-core/src/symbol_analysis/tests.rs (383 LOC) `tests.rs` carries the existing `mod tests` body verbatim — no test content changed; the `use super::*;` makes private items like `nm_range_covers` and `InputSectionIndex` still visible inside the child module. 22/22 tests still pass. Pure refactor; no behavior change to the public API.
…s' into feat/425-rodata-symbol-attribution # Conflicts: # crates/fbuild-build/src/symbol_analyzer.rs
Uh oh!
There was an error while loading. Please reload this page.
zackees added a commit
that referenced
this pull request
Jun 6, 2026
…-S3 (#448) The Phase 2 + Phase 3 acceptance gate from the #434 meta. Exercises the full `fbuild build` → `fbuild bloat` flow against the latest FastLED master + ESP32-S3 toolchain to verify: - build_info.json carries the four #428 toolchain paths (nm_path, cppfilt_path, readelf_path, objdump_path) and mirrors them onto the PIO-shape `aliases` block. - `fbuild bloat .` works zero-flag and lands `report.json` + `report.md` at the documented Phase 3 path `<project>/.fbuild/build/esp32s3/bloat-report/`. - stdout prints both absolute paths on exit (Phase 3 invariant). - Symbol counts, totals, and the >100 map-derived (#427) row count are within bounds generous enough not to flake on FastLED master drift, but tight enough to catch real regressions. - Markdown output renders the top-flash table. Marked `#[ignore]` so ordinary `cargo test` runs aren't network- dependent. Wire into the nightly CI matrix via workflow_dispatch + scheduled weekly to surface FastLED master regressions here too. Closes#442. Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
4 tasks
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 freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
Summary
Closes#425. Adds map-derived synthetic symbols to
fbuild symbolsso anonymous merged rodata blocks (the.rodata.<owner>.str1.<N>and.rodata.<owner>.cst<N>sub-sections that holdFL_WARN/FL_LOGand other compiler-emitted strings) are attributed to the owning function instead of stopping at the object file. Stacked on top of #424.What changed
extract_owner_from_section()strips known prefixes (.text.,.literal.,.rodata.,.data.,.bss.,.data.rel.ro.,.gnu.linkonce.{t,r,d,b}.) and known suffixes (.str1.<N>,.cst<N>) to return the mangled owner of an input section.collect_map_derived_owners()walks map ranges, drops any whose address is already covered by an nm-listed symbol, and returns the survivors with their extracted owner.build_fine_grained_map_with_synth()emits syntheticFineGrainedSymbolrows taggedsource: \"map-derived\".source: \"nm\" | \"map-derived\"field onFineGrainedSymbolwithserde(default)so JSON written by earlierfbuild symbolsstill deserialises.Empirical impact (ESP32-S3 Blink master ELF, the case that motivated #425)
Single largest new attribution: 58 KB of
main.cpp.orodata now correctly named asfl::ClocklessIdf5<3, TIMING_WS2812_800KHZ, GRB, 0, false, 280>::ClocklessIdf5— the FL_WARN string pool pulled in by the NEOPIXEL chipset header. Other meaningful new resolutions:fl::XY(unsigned char, unsigned char)(libFastLED.a/fl.gfx+.cpp.o)printBeforeSetupInfo()(libFrameworkArduino.a/chip-debug-report.cpp.o)fl::ChannelEngineRMTImpl::attemptAllocationRecovery(...)fl::ChannelEngineRMTImpl::reconfigureForNetwork()Test plan
fbuild-core::symbol_analysis::testscovering every recognised input-section shape, the no-double-count invariant when nm AND map both reference an address, range-overlap detection, and end-to-end synthesis throughbuild_fine_grained_map.fbuild-build::symbol_analyzer::testsstill pass (toolchain prefix derivation).uv run --script lintclean on touched files (rustfmt + clippy via fbuild's own gate).Out of scope
.rodatablocks,.flash_rodata_dummyalignment padding) — these stay object-bucketed by design.vtable for ...annotation lines alongside.rodata._ZTV<class>sections; we already produce the right answer because_ZTV<class>demangles cleanly, but a follow-up could short-circuit the c++filt call when the map already gives us the demangled string.🤖 Generated with Claude Code