본문으로 건너뛰기
cccskills
무료GitHub에 공개

android-tombstone-symbolication

Symbolicate the .NET runtime frames in an Android tombstone file. Extracts BuildIds and PC offsets from the native backtrace, downloads debug symbols from the Microsoft symbol server, and runs llvm-symbolizer to produce function names with source file and line numbers. USE FOR triaging a .NET MAUI or Mono Android app crash from a tombstone, resolving native backtrace frames in libmonosgen-2.0.so or libcoreclr.so to .NET runtime source code, or investigating SIGABRT, SIGSEGV, or other native signals originating from the .NET runtime on Android. DO NOT USE FOR pure Java/Kotlin crashes, managed .NET exceptions that are already captured in logcat, or iOS crash logs. INVOKES Symbolicate-Tombstone.ps1 script, llvm-symbolizer, Microsoft symbol server.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

설치 방법 보기

포함된 파일(2)

  • SKILL.md8.0 KB
  • scripts/Symbolicate-Tombstone.ps129.5 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

Android Tombstone .NET Symbolication

Resolves native backtrace frames from .NET Android app crashes (MAUI, Xamarin, Mono) to function names, source files, and line numbers using ELF BuildIds and Microsoft's symbol server.

Inputs: Tombstone file or logcat crash output, llvm-symbolizer (from Android NDK or any LLVM 14+ toolchain), internet access for symbol downloads.

Do not use when: The crash is a managed .NET exception (visible in logcat with a managed stack trace), the crashing library is not a .NET component (e.g., libart.so), or the tombstone is from iOS.


Workflow

Step 1: Parse the Tombstone Backtrace

Each backtrace frame has this format:

#NN pc OFFSET  /path/to/library.so (optional_symbol+0xNN) (BuildId: HEXSTRING)

Extract: frame number, PC offset (hex, already library-relative), library name, and BuildId (32–40 hex chars).

Symbolicate all threads by default (background threads like GC/finalizer often have useful .NET frames). The crashing thread's backtrace is listed first; additional threads appear after --- --- --- markers.

Format notes:

  • The script auto-detects #NN pc frame lines with or without a backtrace: header, and strips logcat timestamp/tag prefixes automatically.
  • Logcat-captured tombstones often omit BuildIds. Recover via adb shell readelf -n, CI build artifacts, or the .NET runtime NuGet package.
  • GitHub issue pastes may mangle #1 pc into issue links — replace org/repo#N pc with #N pc before saving to a file.
  • If the script fails to parse a format, fall back to manual extraction of #NN pc OFFSET library.so (BuildId: HEX) tuples.

Step 2: Identify .NET Runtime Libraries

Filter frames to .NET runtime libraries:

LibraryRuntime
libmonosgen-2.0.soMono (MAUI, Xamarin, interpreter)
libcoreclr.soCoreCLR (JIT mode)
libSystem.*.so.NET BCL native components (Native, Globalization.Native, IO.Compression.Native, Security.Cryptography.Native.OpenSsl, Net.Security.Native)

NativeAOT: No libcoreclr.so or libmonosgen-2.0.so — the runtime is statically linked into the app binary (e.g., libMyApp.so). The libSystem.*.so BCL libraries remain separate and can be symbolicated via the symbol server. For the app binary itself, you need the app's own debug symbols.

Skip libc.so, libart.so, and other Android system libraries unless the user specifically asks.

Step 3: Download Debug Symbols

For each unique .NET BuildId, download debug symbols:

https://msdl.microsoft.com/download/symbols/_.debug/elf-buildid-sym-<BUILDID>/_.debug
curl -sL "https://msdl.microsoft.com/download/symbols/_.debug/elf-buildid-sym-1eb39fc72918c7c6c0c610b79eb3d3d47b2f81be/_.debug" \
  -o libmonosgen-2.0.so.debug

Verify with file libmonosgen-2.0.so.debug — should show ELF 64-bit ... with debug_info, not stripped. If the download returns 404 or HTML, symbols are not published for that build. Do not add or subtract library base addresses — offsets in tombstones are already library-relative.

Step 4: Symbolicate Each Frame

llvm-symbolizer --obj=libmonosgen-2.0.so.debug -f -C 0x222098

Output:

ves_icall_System_Environment_FailFast
/__w/1/s/src/runtime/src/mono/mono/metadata/icall.c:6244

The /__w/1/s/ prefix is the CI workspace root — the meaningful path starts at src/runtime/, mapping to dotnet/dotnet VMR.

Step 5: Present the Symbolicated Backtrace

Combine original frame numbers with resolved function names and source locations:

#00  libc.so              abort+164
#01  libmonosgen-2.0.so   ves_icall_System_Environment_FailFast        (mono/metadata/icall.c:6244)
#02  libmonosgen-2.0.so   do_icall                                     (mono/mini/interp.c:2457)
#03  libmonosgen-2.0.so   mono_interp_exec_method                      (mono/mini/interp.c)

For unresolved frames (??), keep the original line with BuildId and PC offset.

Automation Script

scripts/Symbolicate-Tombstone.ps1 automates the full workflow:

pwsh scripts/Symbolicate-Tombstone.ps1 -TombstoneFile tombstone_01.txt -LlvmSymbolizer llvm-symbolizer

Flags: -CrashingThreadOnly (limit to crashing thread), -OutputFile path (write to file), -ParseOnly (report libraries/BuildIds/URLs without downloading), -SkipVersionLookup (skip runtime version identification).


Finding llvm-symbolizer

Check the Android NDK first: $ANDROID_NDK_ROOT/toolchains/llvm/prebuilt/*/bin/llvm-symbolizer or $ANDROID_HOME/ndk/*/toolchains/llvm/prebuilt/*/bin/llvm-symbolizer. Also available via brew install llvm, apt install llvm, or xcrun --find llvm-symbolizer on macOS.

If unavailable, complete steps 1–3 and present the download commands and llvm-symbolizer commands for the user to run. Do not spend time installing LLVM.


Understanding the Output

CI source paths use these prefixes:

Path prefixMaps to
/__w/1/s/src/runtime/src/runtime/ in dotnet/dotnet VMR
/__w/1/s/src/mono/src/mono/ in the VMR (older builds)
/__w/1/s/VMR root

Runtime Version Identification

The script identifies the exact .NET runtime version by matching BuildIds against locally-installed runtime packs. It searches: SDK packs ($DOTNET_ROOT/packs/), NuGet cache (~/.nuget/packages/), and NuGet.org as an online fallback. When found, it extracts the version and source commit from the .nuspec <repository commit="..." /> element. Pass -SkipVersionLookup to disable. Requires llvm-readelf (auto-discovered from the NDK).


Validation

  1. file <debug-file> shows ELF ... with debug_info, not stripped
  2. At least one .NET frame resolves to a function name (not ??)
  3. Resolved paths contain recognizable .NET runtime structure (e.g., mono/metadata/, mono/mini/)

Stop Signals

  • No .NET frames found: Report parsed frames and stop.
  • All frames resolved: Present symbolicated backtrace. Do not trace into source or attempt to build/debug the runtime.
  • Symbols not available (404): One attempt per BuildId, then stop. Report unsymbolicated frames with BuildIds and offsets.
  • llvm-symbolizer not available: Use -ParseOnly, present manual commands. Do not install LLVM.

Common Pitfalls

  • Missing BuildIds: Logcat tombstones often omit BuildIds. Recover via: adb shell readelf -n /path/to/lib.so, CI build artifacts, or the runtime NuGet package (~/.dotnet/packs/Microsoft.NETCore.App.Runtime.Mono.android-arm64/<version>/). Prefer pulling raw tombstone files (adb shell cat /data/tombstones/tombstone_XX) which always include BuildIds.
  • Symbols not found (404): Pre-release/internal builds may not publish symbols. Check for local unstripped .so/.so.dbg in build artifacts or the NuGet runtime pack.
  • NativeAOT: No runtime .so in the tombstone — runtime is in the app binary. libSystem.*.so BCL libraries still work with the symbol server; the app binary needs its own debug symbols.
  • Wrong llvm-symbolizer version: Use LLVM 14+ for best DWARF compatibility.
  • Multiple BuildIds: Each .NET library has its own BuildId — download symbols for each separately.

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

12-agent academic paper writing pipeline. 10 modes (full/plan/outline/revision/revision-coach/abstract/lit-review/format-convert/citation-check/disclosure). 6 paper types, 5 citation formats, bilingual abstracts, LaTeX/DOCX-via-Pandoc/PDF output. Style Calibration + Writing Quality Check + Anti-Patterns with IRON RULE markers. Triggers: write paper, academic paper, guide my paper, parse reviews, AI disclosure, 寫論文, 學術論文, 引導我寫論文, 審查意見.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

bouclem/skills72026년 5월 31일 업데이트

Multi-perspective academic paper review with dynamic reviewer personas. Simulates 5 independent reviewers (EIC + 3 peer reviewers + Devil's Advocate) with field-specific expertise. Supports full review, re-review (verification), quick assessment, methodology focus, Socratic guided, and calibration modes. Triggers on: review paper, peer review, manuscript review, referee report, review my paper, critique paper, simulate review, editorial review, calibrate reviewer, reviewer calibration, measure reviewer accuracy.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

bouclem/skills72026년 5월 31일 업데이트

Orchestrator for the full academic research pipeline: research -> write -> integrity check -> review -> revise -> re-review -> re-revise -> final integrity check -> finalize. Coordinates deep-research, academic-paper, and academic-paper-reviewer into a seamless 10-stage workflow with mandatory integrity verification, two-stage peer review, and reproducible quality gates. Triggers on: academic pipeline, research to paper, full paper workflow, paper pipeline, end-to-end paper, research-to-publication, complete paper workflow.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

bouclem/skills72026년 5월 31일 업데이트

Create, iterate, and scale paid ad creative for Google Ads, Meta, LinkedIn, TikTok, and similar platforms. Use when generating headlines, descriptions, primary text, or large sets of ad variations for testing and performance optimization.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

bouclem/skills72026년 5월 31일 업데이트

This skill should be used when the user asks to "implement LLM-as-judge", "compare model outputs", "create evaluation rubrics", "mitigate evaluation bias", or mentions direct scoring, pairwise comparison, position bias, evaluation pipelines, or automated quality assessment.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

bouclem/skills72026년 5월 31일 업데이트

AI agent development workflow for building autonomous agents, multi-agent systems, and agent orchestration with CrewAI, LangGraph, and custom agents.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

bouclem/skills72026년 5월 31일 업데이트

bouclem의 스킬 모두 보기

이 스킬의 문제 신고