Apps/integration 0.1.9

.gitattributes

@@ -29,3 +29,7 @@
 *.jpg       binary
 *.pdf       binary
 *.sqlite    binary
+
+# A torture fixture whose CRLF line endings ARE the trap it sets (a BOM and CRLF in one
+# page); normalising it would remove the case the suite asserts.
+parser/src/test-data/web/torture/quirks.html -text

.github/workflows/build-engines.yml

@@ -73,7 +73,9 @@ jobs:
             id="$(bash graph/pipeline/run-souffle.sh --language "$lang" --print-engine-id)"
             echo "$lang: $id"; printf '%s' "$id" > "gen/$lang.id"
             bash graph/pipeline/run-souffle.sh --language "$lang" --emit-program "gen/$lang.dl"
-            souffle -I graph -g "gen/$lang.cpp" "gen/$lang.dl" 2> "gen/$lang.gen.log" || { cat "gen/$lang.gen.log"; exit 1; }
+            # -j makes souffle EMIT the parallel loops (without it the C++ has none); whether a
+            # platform's binary runs them in parallel is decided at compile time below.
+            souffle -I graph -j 8 -g "gen/$lang.cpp" "gen/$lang.dl" 2> "gen/$lang.gen.log" || { cat "gen/$lang.gen.log"; exit 1; }
             awk '/No rules\/facts defined/{skip=2;next} skip>0{skip--;next} {print}' "gen/$lang.gen.log"
           done
           # the query programs: self-contained (no #include), keyed by dl_program.py itself
@@ -89,6 +91,33 @@ jobs:
           bash .github/scripts/query-smoke.sh expect gen/queries
           cp .github/scripts/query-smoke.sh gen/queries/smoke.sh
           cp -r /usr/include/souffle gen/souffle
+          # the seqlock fix (see souffle_overlay in run-souffle.sh): the write-entry RMW
+          # must be seq_cst or a weakly-ordered CPU lets data stores pass the version-odd
+          # store and readers validate garbage. Patched here so every platform's binary
+          # is built from the same fixed headers; the engine id carries +seqlock-fix-4,
+          # so these headers must carry the SAME patches run-souffle's overlay applies —
+          # a binary labeled fix-4 but built from lesser headers is the exact mislabeling
+          # the salt exists to prevent.
+          # the header layout differs by install (deb: souffle/utility; brew: souffle/souffle/utility),
+          # so the patched files are FOUND, not assumed — a miss fails the build here, loudly
+          PU=$(find gen/souffle -name ParallelUtil.h | head -1); [ -n "$PU" ]
+          sed -i 's/version\.fetch_or(0x1, std::memory_order_acquire)/version.fetch_or(0x1, std::memory_order_seq_cst)/g' "$PU"
+          grep -q 'fetch_or(0x1, std::memory_order_seq_cst)' "$PU"
+          ! grep -q 'fetch_or(0x1, std::memory_order_acquire)' "$PU"
+          # fix-3: the publication fences, applied by the same anchored patch the local
+          # overlay uses (extracted from run-souffle.sh so the two cannot drift)
+          BT=$(find gen/souffle -name BTree.h | head -1); [ -n "$BT" ]
+          awk '/^import glob, os, sys$/,/^PYEOF$/' graph/pipeline/run-souffle.sh | sed '$d' > /tmp/fences.py
+          python3 /tmp/fences.py "$BT"
+          [ "$(grep -c 'seqlock-fix-3' "$BT")" = "4" ]
+          # fix-4: the speculative-fetch guards, applied by the same extracted patch
+          FW=$(find gen/souffle -name ConcurrentFlyweight.h | head -1); [ -n "$FW" ]
+          RT=$(find gen/souffle -name RecordTableImpl.h | head -1); [ -n "$RT" ]
+          [ "$(grep -c 'seqlock-fix-4' "$FW")" = "1" ]
+          [ "$(grep -c 'seqlock-fix-4' "$RT")" = "1" ]
+          [ "$(grep -c 'seqlock-fix-4' "$BT")" = "1" ]
+          BD=$(find gen/souffle -name BTreeDelete.h | head -1); [ -n "$BD" ]
+          [ "$(grep -c 'seqlock-fix-4' "$BD")" = "1" ]
           # one key for the whole set; a partial match restores the previous set
           echo "key=$(cat gen/*.id gen/queries/*.id | sha256sum | cut -c1-16)" >> "$GITHUB_OUTPUT"
           # The compile flags live in THIS file and ENGINE_ID does not cover them, so its
@@ -135,7 +164,15 @@ jobs:
           for lang in $LANGUAGES; do
             if cmp -s "gen/$lang.id" "engines/$lang/ENGINE_ID"; then echo "$lang: cached, rules unchanged"; continue; fi
             mkdir -p "engines/$lang"
-            c++ -std=c++17 -O3 -w -static-libstdc++ -static-libgcc -I gen "gen/$lang.cpp" -o "engines/$lang/axiomcode-engine-$lang"
+            # OpenMP on BOTH architectures: the arm64 crash mode was two weak-ordering
+            # holes in the vendored headers (write-entry RMW + unfenced node publication),
+            # both patched above and validated 10/10 under load on linux-arm64 and
+            # darwin-arm64 (seqlock-fix-3). libgomp links STATICALLY so the binary runs on
+            # machines with no gcc runtime. The .parallel marker beside the binary is what
+            # run-souffle.sh reads to pass a real -j at run time.
+            OMP="-fopenmp -Wl,-Bstatic,-lgomp,-Bdynamic"
+            c++ -std=c++17 -O3 -w $OMP -static-libstdc++ -static-libgcc -I gen "gen/$lang.cpp" -o "engines/$lang/axiomcode-engine-$lang"
+            [ -n "$OMP" ] && touch "engines/$lang/axiomcode-engine-$lang.parallel"
             cp "gen/$lang.id" "engines/$lang/ENGINE_ID"
           done
           mkdir -p engines/queries
@@ -169,8 +206,9 @@ jobs:
             fc /b gen\%%L.id engines\%%L\ENGINE_ID >nul 2>&1
             if errorlevel 1 (
               if not exist engines\%%L mkdir engines\%%L
-              cl /nologo /std:c++17 /O2 /EHsc /bigobj /w /permissive- /Zc:__cplusplus /D_CRT_SECURE_NO_WARNINGS /DNOMINMAX /DUSE_CUSTOM_GETOPTLONG /I gen gen\%%L.cpp /Fe:engines\%%L\axiomcode-engine-%%L.exe
+              cl /nologo /std:c++17 /O2 /EHsc /bigobj /w /permissive- /Zc:__cplusplus /D_CRT_SECURE_NO_WARNINGS /DNOMINMAX /DUSE_CUSTOM_GETOPTLONG /openmp /I gen gen\%%L.cpp /Fe:engines\%%L\axiomcode-engine-%%L.exe
               if errorlevel 1 exit /b 1
+              type nul > engines\%%L\axiomcode-engine-%%L.exe.parallel
               copy /y gen\%%L.id engines\%%L\ENGINE_ID
             ) else (
               echo %%L: cached, rules unchanged

.github/workflows/ci.yml

@@ -52,7 +52,7 @@ permissions:
   contents: read
 
 env:
-  # bin/axiomcode and the parser need Node ≥ 22.5.
+  # bin/axiomcode and the parser need Node ≥ 22.13 (node:sqlite without a flag).
   NODE_VERSION: '22'
   SOUFFLE_VERSION: '2.5'
   SOUFFLE_SHA512: '6b86e554f6aa5abf8a8b55d8312ae37c0957c5bd6c9edeea89246db9406f645ec5e600b84fe6636b1c163da556f0da6c3d2dad46c1083413f2fcf4f95b9ac62c'
@@ -371,6 +371,16 @@ jobs:
         if: matrix.lang == 'csharp'
         run: dotnet build -c Release graph/test/csharp/ground-truth/AxiomCsOracle
 
+      # `--library auto` decompiles a dependency that ships no sources jar; the Java query case for it needs Vineflower
+      - name: fetch the decompiler `--library auto` uses for Java class jars
+        if: matrix.lang == 'java'
+        run: mvn -q dependency:get -Dartifact=org.vineflower:vineflower:1.10.1
+
+      # `--library auto` compiles a NuGet package by decompiling it; the query case for it needs the decompiler
+      - name: install the decompiler `--library auto` uses for NuGet packages
+        if: matrix.lang == 'csharp'
+        run: dotnet tool install -g ilspycmd --version 8.2.0.7535
+
       - name: cache the Soufflé package
         uses: actions/cache@v4
         with:

README.md

@@ -43,7 +43,7 @@
   <a href="https://www.npmjs.com/package/@axiomcode/code-graph"><img alt="npm" src="https://img.shields.io/npm/v/@axiomcode/code-graph?label=npm"></a>
   <a href="https://github.com/AxiomCodeAI/axiomcodegraph/actions/workflows/nightly.yml"><img alt="Nightly (dev)" src="https://github.com/AxiomCodeAI/axiomcodegraph/actions/workflows/nightly.yml/badge.svg?branch=dev"></a>
   <a href="LICENSE.md"><img alt="License: FSL-1.1-Apache-2.0" src="https://img.shields.io/badge/license-FSL--1.1--Apache--2.0-blue"></a>
-  <img alt="Node ≥ 22.5" src="https://img.shields.io/badge/node-%E2%89%A5%2022.5-brightgreen">
+  <img alt="Node ≥ 22.13" src="https://img.shields.io/badge/node-%E2%89%A5%2022.13-brightgreen">
 </p>
 
 <p align="center">
@@ -125,7 +125,7 @@ AxiomCode Graph is two parts. The **engine** (`@axiomcode/code-graph` on npm) pa
 graph; it also provides the `axiomcode` command and an MCP server. The **plugin** (`plugins/axiomcode/`) is the
 agent-facing frontend: a skill, four MCP tools, and hooks. Install the engine first.
 
-Requirements: **Node ≥ 22.5** and **Python 3** (`python3`, or `python` / `py` on Windows). On Windows, also
+Requirements: **Node ≥ 22.13** and **Python 3** (`python3`, or `python` / `py` on Windows). On Windows, also
 [Git for Windows](https://git-scm.com/download/win): the CLI runs under its bash. The engine ships as a prebuilt
 binary for macOS (Apple Silicon and Intel), Linux (x64 and arm64) and Windows x64, and `npm install` takes the one for
 your platform. No Soufflé and no compiler are needed, with one exception:
@@ -255,8 +255,8 @@ looked up again in the graph before you see it; the `verified:` line is that che
 
 ### Support for agents
 
-Every agent below gets the four MCP tools and the skill; the hooks, which add the graph's edges to the agent's
-own file reads and searches, run where the last column says so.
+Every agent below gets the four MCP tools and the skill; the hooks, which keep the graph current and report what an
+edit breaks, run where the last column says so. No hook annotates the agent's own reads and searches.
 
 | Agent | Install | Uninstall | Hooks |
 |---|---|---|---|
@@ -308,12 +308,12 @@ about one change had to read 95 of 43,793 methods, and every true direct caller
 
 ## CLI commands
 
-Four questions, each answered as numbered places with the code of the function each one sits in. The MCP server
-offers the same four as tools: `find(question)`, `impact(name)`, `path(start, end)` and `tests()`.
+Three questions, each answered as numbered places with the code of the function each one sits in. The MCP server
+offers the same three as tools: `impact(name)`, `path(start, end)` and `tests()`. Finding where code lives is
+left to your own search: bring the name you found to these commands.
 
 | command | what it answers |
 |---|---|
-| `axiomcode find "<question>"` | where the code for a task lives, when you have it in words and not yet a name |
 | `axiomcode impact <name>` | who calls it, what a change to it reaches, and the tests that exercise it |
 | `axiomcode impact` | the same for the declarations your uncommitted edits changed; the answer starts with `your edits:` |
 | `axiomcode path <A> <B>` | how A reaches B: every hop of the call chain, with the code at each call |
@@ -334,9 +334,9 @@ command's usage.
 The graph stays current on its own. Every file the parser reads is recorded with its hash at build time; after an
 edit, a shell command, a finished turn, at session start, and before a query, anything that differs starts one
 background rebuild per repository, with the language, `--src` and `--library` of the graph it replaces. Every
-command keeps reading the previous graph until the new one is indexed and swapped in. A query waits up to
-`AXIOMCODE_FRESH_WAIT` seconds (default 10) for it, then answers from the previous graph with a `graph refresh:` line
-naming the files it predates. The MCP server also checks every repository it has answered for once 15 minutes have
+command keeps reading the previous graph until the new one is indexed and swapped in. A query answers from the
+previous graph at once, with a `graph refresh:` line naming the files it predates and the rows in them marked;
+`--fresh` waits for the rebuild instead, and `AXIOMCODE_FRESH_WAIT` (seconds, default 0) lets every query wait that long. The MCP server also checks every repository it has answered for once 15 minutes have
 passed since its last update (`AXIOMCODE_REFRESH_INTERVAL`, seconds; 0 turns it off), which catches edits made while
 a session sits idle. The graph records when and why it was built in `index_meta` (`refreshed_at`, `refresh_reason`).
 `AXIOMCODE_NO_REFRESH=1` turns the rebuilds off, not the check: an answer from a graph older than an edit still