bug for parse c++

[chen135790]

Version

codebase-memory-mcp v0.8.1

Platform

Linux (x64)

Install channel

GitHub release archive / install.sh / install.ps1

Binary variant

standard

What happened, and what did you expect?

When a C++ file contains #ifdef / #else / #endif blocks around function signatures, the tree-sitter C++ parser silently skips all code from that point until a much later line, resulting in massive gaps in the knowledge graph.

Reproduction

File: SurfaceFlinger.cpp (11,657 lines)

Trigger pattern (line 2064):

#ifdef FLYME_GRAPHICS_EXTEND_LUMARGB
status_t SurfaceFlinger::addRegionSamplingListener(const Rect& samplingArea,
                                                   const sp<IBinder>& stopLayerHandle,
                                                   const sp<IRegionSamplingListener>& listener,
                                                   const bool rgbSample) {
#else
status_t SurfaceFlinger::addRegionSamplingListener(const Rect& samplingArea,
                                                   const sp<IBinder>& stopLayerHandle,
                                                   const sp<IRegionSamplingListener>& listener) {
#220 

### Logs

```text
## Bug Description

When a C++ file contains `#ifdef` / `#else` / `#endif` blocks around function signatures, the tree-sitter C++ parser silently skips all code from that point until a much later line, resulting in massive gaps in the knowledge graph.

## Reproduction

**File:** `SurfaceFlinger.cpp` (11,657 lines)

**Trigger pattern (line 2064):**

#ifdef FLYME_GRAPHICS_EXTEND_LUMARGB
status_t SurfaceFlinger::addRegionSamplingListener(const Rect& samplingArea,
                                                   const sp<IBinder>& stopLayerHandle,
                                                   const sp<IRegionSamplingListener>& listener,
                                                   const bool rgbSample) {
#else
status_t SurfaceFlinger::addRegionSamplingListener(const Rect& samplingArea,
                                                   const sp<IBinder>& stopLayerHandle,
                                                   const sp<IRegionSamplingListener>& listener) {
#endif
Result:
- Methods indexed: lines 463–2065, then 10619–11637
- Gap: lines 2066–10618 (~8,552 lines) completely missing from graph
- Critical methods SurfaceFlinger::commit (line 2758) and SurfaceFlinger::composite (line 3007) are not indexed
Expected Behavior
All methods in the file should be indexed regardless of #ifdef blocks around signatures.
Actual Behavior
The parser appears to bail when encountering duplicate function signatures in #ifdef/#else branches, skipping everything until it finds a "clean" parse point ~8,500 lines later.
Impact
This affects any C/C++ codebase using conditional compilation for platform-specific or feature-flag variants of function signatures — very common in Android/AOSP, Linux kernel, and embedded projects.
Environment
- codebase-memory-mcp v0.8.1
- Linux amd64
- Indexing mode: full / moderate / fast (all exhibit the same gap)

Diagnostics trajectory (memory / performance / leak issues)

**File:** `SurfaceFlinger.cpp` (11,657 lines)

**Trigger pattern (line 2064):**

#ifdef FLYME_GRAPHICS_EXTEND_LUMARGB
status_t SurfaceFlinger::addRegionSamplingListener(const Rect& samplingArea,
                                                   const sp<IBinder>& stopLayerHandle,
                                                   const sp<IRegionSamplingListener>& listener,
                                                   const bool rgbSample) {
#else
status_t SurfaceFlinger::addRegionSamplingListener(const Rect& samplingArea,
                                                   const sp<IBinder>& stopLayerHandle,
                                                   const sp<IRegionSamplingListener>& listener) {
#endif

Project scale (if relevant)

No response

Confirmations

• I searched existing issues and this is not a duplicate.
• My reproduction uses shareable code (a dummy snippet or a public OSS repository), not proprietary code.

[blankanswer]

Hi, I can take a look at this~

My plan is to keep the first PR scoped and reproducible:

• add a minimal regression fixture for a C++ method declaration/definition split by #ifdef / #else / #endif around the signature
• confirm whether the current extractor drops the methods after the conditional block
• improve the C++ extraction fallback/recovery path so later methods in the file are still indexed instead of losing a large tail range
• keep the fix focused on indexing continuity rather than trying to fully model all preprocessor branches

I’ll first reduce the SurfaceFlinger.cpp case into a small shareable fixture, then open a PR with the failing test and the minimal recovery fix.

[DeusData]

Thanks for the compact C++ repro. Triage: this looks like a parser recovery bug around preprocessor-gated method signatures. #949 is open for the fix; review should make sure it recovers the callable without duplicating or mis-owning definitions.

[DeusData]

Fixed on main by 65a176e (PR #1050, shipping in v0.9.1) — this is the same root-cause class as #961: both #ifdef branches are present in the raw parse, so the duplicated opening brace makes the region unbalanced, tree-sitter emits an ERROR region, and the defs walk dropped everything it covered.

The fix recovers definitions from the preprocessed second pass (simplecpp evaluates the conditionals so exactly one signature branch survives, and same-file token lines stay aligned, so recovered defs keep correct line numbers), gated to defs that intersect a raw ERROR region and are otherwise absent.

Verified on your exact SurfaceFlinger shape (signature split across #ifdef/#else with a shared body):

function	before	after
addRegionSamplingListener (the split one)	dropped	indexed, lines 13–21
removeRegionSamplingListener (after the split)	dropped ("skips all code")	indexed
later functions	dropped	indexed

Could you retest against a main build (or v0.9.1 when it ships) on the full 11k-line SurfaceFlinger.cpp and confirm the gap closes there too? If any residual region stays missing, the parse_partial coverage signal (index_status) will name it — paste that output and we'll chase the remainder.

[github-actions]

Thanks for the report! To move this forward we need a bit more so we can reproduce it ourselves:

• the codebase-memory-mcp --version you're on
• the exact steps or command you ran
• a public repo (or a small dummy snippet) that shows the problem — please don't paste proprietary code

Once that's here we'll pick it straight back up. Heads-up: issues left awaiting-reporter are automatically closed after a few weeks of silence, but a comment reopens the door anytime.

[DeusData]

Follow-up: the C++ #ifdef-split recovery this issue reported (fixed by #1050) was further hardened by #949 (merged ef14822), which adds explicit expanded→original line remapping so recovered definitions land on correct lines even when resolved #includes shift the preprocessed line numbers — a case the original recovery approximated. Your SurfaceFlinger shape stays covered; still awaiting your retest confirmation on a v0.9.1 build against the full 11k-line file.

[github-actions]

This issue has been waiting on more information for 21 days (a version, exact steps, or a public repro), so it's now marked stale. It will not be closed automatically — a maintainer reviews stale issues by hand. Adding a comment with the details clears the label and puts it straight back in the queue. We're happy to pick it back up the moment we can reproduce it.