mirror of
https://github.com/DeusData/codebase-memory-mcp.git
synced 2026-10-19 05:40:03 +08:00
Martin Vogel
81264814f1
Answers a question the existing tooling could not: of everything we
allocate, how much is never used?
Lanes, all driven by scripts/memwaste.sh:
event every allocation observed through the memory core — size, age,
churn, and whether the bytes were ever written
access SanitizerCoverage load/store instrumentation, so "allocated and
never touched" is measured rather than inferred
scaling k vs 2k replicas through clang profile counters, to catch
super-linear work per file
padding allocator slack per size class
ratchet the numbers recorded per venue so a regression is visible
The event lane needed a foreign-free counter: on Windows the override sees
frees for blocks it never allocated (they come from the CRT), and counting
those as untracked tripped the soundness gate. mem_events now separates
them (14 daemon / 5 worker on the Go corpus).
Two portability fixes the lanes surfaced, both real:
- the enable decision must not be cached before `environ` exists. Under
-fsanitize-coverage the UBSan preinit runs dlsym -> malloc -> our
wrapper before glibc sets up environ, so the lane silently disabled
itself and produced no dump at all on Linux.
- the access map needs 48 bits: arm64 Linux hands out addresses above
2^47, which a 47-bit map dropped on the floor.
mem_profile is retired into this: it measured a strict subset and had its
own sampling story.
Local-only by decision — not a PR-CI gate. _memwaste.yml runs the lanes on
all three platforms report-only, and pr.yml/release.yml carry the wiring.
Signed-off-by: Martin Vogel <martin.vogel.tech@gmail.com>
127 lines
8.6 KiB
JSON
127 lines
8.6 KiB
JSON
{
|
|
"about": "Defect class x platform coverage of the sanitizer lanes. Every cell is covered, partial (with why), or a named limit (with why). Validated by scripts/sanitizer-coverage-map.py map. Which functions each lane actually EXECUTES is the run/union/ratchet part of the same tool.",
|
|
"platforms": ["linux-x64", "linux-arm64", "macos-arm64", "windows-x64", "windows-arm64"],
|
|
"classes": [
|
|
{
|
|
"name": "heap/stack memory errors",
|
|
"cells": {
|
|
"linux-x64": {"status": "covered", "lane": "asan", "venue": "PR CI"},
|
|
"linux-arm64": {"status": "covered", "lane": "asan", "venue": "local container"},
|
|
"macos-arm64": {"status": "partial", "lane": "asan+diag", "why": "use-after-return and use-after-scope only on the Homebrew-LLVM diag lane, which is local-only"},
|
|
"windows-x64": {"status": "covered", "lane": "asan", "venue": "PR CI"},
|
|
"windows-arm64": {"status": "limit", "lane": "none", "why": "no ASan runtime for Windows ARM64; the class is covered for Windows by the x64 leg"}
|
|
}
|
|
},
|
|
{
|
|
"name": "allocator-path memory errors",
|
|
"cells": {
|
|
"linux-x64": {"status": "partial", "lane": "asan", "why": "test builds use the C runtime allocator: mimalloc bindings, the --wrap shims and the slab never run sanitized; mimalloc itself is built without ASan"},
|
|
"linux-arm64": {"status": "partial", "lane": "asan", "why": "same as linux-x64"},
|
|
"macos-arm64": {"status": "partial", "lane": "asan", "why": "same as linux-x64; the zone-observer path runs only in the memwaste flavours"},
|
|
"windows-x64": {"status": "partial", "lane": "asan", "why": "same as linux-x64"},
|
|
"windows-arm64": {"status": "limit", "lane": "none", "why": "no ASan runtime for Windows ARM64"}
|
|
}
|
|
},
|
|
{
|
|
"name": "undefined behaviour",
|
|
"cells": {
|
|
"linux-x64": {"status": "partial", "lane": "ubsan", "why": "tre (regex) is built with -fno-sanitize=alignment"},
|
|
"linux-arm64": {"status": "partial", "lane": "ubsan", "why": "tre (regex) is built with -fno-sanitize=alignment"},
|
|
"macos-arm64": {"status": "partial", "lane": "ubsan", "why": "tre (regex) is built with -fno-sanitize=alignment"},
|
|
"windows-x64": {"status": "partial", "lane": "ubsan", "why": "tre (regex) is built with -fno-sanitize=alignment"},
|
|
"windows-arm64": {"status": "partial", "lane": "ubsan-trap", "why": "trap mode: a violation stops the run without a message"}
|
|
}
|
|
},
|
|
{
|
|
"name": "leaks",
|
|
"cells": {
|
|
"linux-x64": {"status": "covered", "lane": "lsan", "venue": "PR CI (on by default under ASan)"},
|
|
"linux-arm64": {"status": "covered", "lane": "lsan", "venue": "local container"},
|
|
"macos-arm64": {"status": "covered", "lane": "test-lsan", "venue": "local (Homebrew LLVM; Apple clang refuses detect_leaks)"},
|
|
"windows-x64": {"status": "partial", "lane": "memwaste-event", "why": "no leak sanitizer on Windows; the waste lane's live-per-site report at the last phase and the soak slope stand in"},
|
|
"windows-arm64": {"status": "partial", "lane": "memwaste-event", "why": "no leak sanitizer on Windows; the waste lane's live-per-site report at the last phase and the soak slope stand in"}
|
|
}
|
|
},
|
|
{
|
|
"name": "uninitialised reads",
|
|
"cells": {
|
|
"linux-x64": {"status": "partial", "lane": "msan", "why": "CI only; deep-recursion suites need larger thread stacks under MSan"},
|
|
"linux-arm64": {"status": "partial", "lane": "msan", "why": "locally red: three deep-recursion suites overflow their thread stacks"},
|
|
"macos-arm64": {"status": "partial", "lane": "memwaste-access", "why": "no MSan on macOS; the access lane's read-before-write verdict on a real corpus is the second net, per block, not per byte"},
|
|
"windows-x64": {"status": "partial", "lane": "memwaste-access", "why": "no MSan on Windows; the access lane (clang) is the second net"},
|
|
"windows-arm64": {"status": "partial", "lane": "memwaste-access", "why": "no MSan on Windows; the access lane (clang) is the second net"}
|
|
}
|
|
},
|
|
{
|
|
"name": "data races",
|
|
"cells": {
|
|
"linux-x64": {"status": "partial", "lane": "tsan", "why": "a subset of suites runs under TSan"},
|
|
"linux-arm64": {"status": "partial", "lane": "tsan", "why": "a subset of suites runs under TSan"},
|
|
"macos-arm64": {"status": "partial", "lane": "tsan", "why": "a subset of suites runs under TSan"},
|
|
"windows-x64": {"status": "limit", "lane": "none", "why": "no TSan on Windows"},
|
|
"windows-arm64": {"status": "limit", "lane": "none", "why": "no TSan on Windows"}
|
|
}
|
|
},
|
|
{
|
|
"name": "static analysis",
|
|
"cells": {
|
|
"linux-x64": {"status": "partial", "lane": "cppcheck+lint-mem+codeql", "why": "vendored trees are filtered out of cppcheck and the analyzer gate"},
|
|
"linux-arm64": {"status": "partial", "lane": "cppcheck+lint-mem", "why": "vendored trees are filtered out"},
|
|
"macos-arm64": {"status": "partial", "lane": "lint-mem", "why": "vendored trees are filtered out"},
|
|
"windows-x64": {"status": "limit", "lane": "none", "why": "static lanes are platform-independent source passes; Windows-only code is analysed by CodeQL only"},
|
|
"windows-arm64": {"status": "limit", "lane": "none", "why": "static lanes are platform-independent source passes"}
|
|
}
|
|
},
|
|
{
|
|
"name": "hostile input",
|
|
"cells": {
|
|
"linux-x64": {"status": "partial", "lane": "security-fuzz+libfuzzer", "why": "fixed MCP payloads in CI; libFuzzer targets for extractors, Cypher and config editors run in dry-run/nightly"},
|
|
"linux-arm64": {"status": "partial", "lane": "libfuzzer", "why": "local only"},
|
|
"macos-arm64": {"status": "partial", "lane": "libfuzzer", "why": "local only (Homebrew LLVM ships libFuzzer; Apple clang does not)"},
|
|
"windows-x64": {"status": "limit", "lane": "none", "why": "libFuzzer is not built for the MinGW toolchain"},
|
|
"windows-arm64": {"status": "limit", "lane": "none", "why": "libFuzzer is not built for the MinGW toolchain"}
|
|
}
|
|
},
|
|
{
|
|
"name": "algorithmic blow-up",
|
|
"cells": {
|
|
"linux-x64": {"status": "covered", "lane": "complexity+memwaste-scaling"},
|
|
"linux-arm64": {"status": "covered", "lane": "complexity+memwaste-scaling"},
|
|
"macos-arm64": {"status": "covered", "lane": "complexity+memwaste-scaling"},
|
|
"windows-x64": {"status": "partial", "lane": "complexity", "why": "the per-function scaling lane needs clang's profile runtime; whole-pipeline counters only"},
|
|
"windows-arm64": {"status": "partial", "lane": "complexity", "why": "the per-function scaling lane needs clang's profile runtime; whole-pipeline counters only"}
|
|
}
|
|
},
|
|
{
|
|
"name": "memory waste",
|
|
"cells": {
|
|
"linux-x64": {"status": "covered", "lane": "memwaste-event+access"},
|
|
"linux-arm64": {"status": "covered", "lane": "memwaste-event+access"},
|
|
"macos-arm64": {"status": "covered", "lane": "memwaste-event+access"},
|
|
"windows-x64": {"status": "partial", "lane": "memwaste-event", "why": "libc wrappers cover mem*/str* and FILE I/O only; the access lane needs a clang build"},
|
|
"windows-arm64": {"status": "partial", "lane": "memwaste-event", "why": "libc wrappers cover mem*/str* and FILE I/O only"}
|
|
}
|
|
},
|
|
{
|
|
"name": "CPU waste",
|
|
"cells": {
|
|
"linux-x64": {"status": "covered", "lane": "memwaste-cpu+scaling"},
|
|
"linux-arm64": {"status": "covered", "lane": "memwaste-cpu+scaling"},
|
|
"macos-arm64": {"status": "covered", "lane": "memwaste-cpu+scaling"},
|
|
"windows-x64": {"status": "partial", "lane": "memwaste-cpu", "why": "no system-call or pthread lock wrappers on Windows: Win32 file and lock APIs are imports --wrap cannot redirect"},
|
|
"windows-arm64": {"status": "partial", "lane": "memwaste-cpu", "why": "no system-call or pthread lock wrappers on Windows: Win32 file and lock APIs are imports --wrap cannot redirect"}
|
|
}
|
|
},
|
|
{
|
|
"name": "executed-code coverage of the lanes",
|
|
"cells": {
|
|
"linux-x64": {"status": "partial", "lane": "sanitizer-coverage-map", "why": "needs a clang job; the gcc legs are not measured"},
|
|
"linux-arm64": {"status": "covered", "lane": "sanitizer-coverage-map"},
|
|
"macos-arm64": {"status": "covered", "lane": "sanitizer-coverage-map"},
|
|
"windows-x64": {"status": "limit", "lane": "none", "why": "no sanitizer lane on Windows executes under clang source coverage yet"},
|
|
"windows-arm64": {"status": "limit", "lane": "none", "why": "no sanitizer lane on Windows executes under clang source coverage yet"}
|
|
}
|
|
}
|
|
]
|
|
}
|