supercov 4.22.0 → 4.23.0

package/docs/cli.md

@@ -921,6 +921,7 @@ terminal or offline environment after the package has been downloaded.
 | Variable | Use |
 | --- | --- |
 | `SUPERCOV_SOURCE_ROOTS` | Comma-separated directories or files that hold your own code, in any language; everything else is left out |
+| `SUPERCOV_TEST_ROOTS` | For an Ant build whose `<batchtest>` directories cannot be read from its files: comma-separated directories that hold its tests |
 | `SUPERCOV_TEST_KIND` | Label the wrapped command as a test level such as `unit` or `e2e` |
 | `SUPERCOV_SOURCE_TOOLS` | Comma-separated names of tools the test command runs that read source as text, such as a linter or a spell checker, besides Biome, oxlint, dprint, cspell and knip. Each is run on your source as you wrote it instead of on the instrumented copy |
 | `SUPERCOV_RUN_ID` | Set by Supercov, not by you: in every process a measured command starts, the run's id, so a test or harness can tell it is measured |

package/docs/supported-suites.md

@@ -736,6 +736,9 @@ npx supercov -- npx playwright test
 | Spock | Exact per feature, under the names Spock itself reports | — |
 | TestNG | Exact per test, each data-provider invocation its own | — |
 | JUnit 4 alone | Exact per test, through Vintage — see below | Maven or Gradle |
+| JUnit 4 under Ant's `<junit>` | Exact per test, under the names Ant's XML report uses | Ant 1.10 — see below |
+| Any platform engine under Ant's `<junitlauncher>` | Exact per test, through the platform listener | Ant 1.10 — see below |
+| TestNG under Ant's `<testng>` | Exact per test, through the TestNG listener | TestNG 6.x with its Ant task — see below |
 
 Multi-module builds are measured module by module: each compiles its own source
 set and forks its own JVM, so each gets a runtime and records evidence of its
@@ -798,6 +801,70 @@ npx supercov -- ./gradlew test
 npx supercov -- ./mvnw verify
 ```
 
+### Ant and Ivy
+
+A project with a `build.xml` (and no Maven or Gradle build) is an Ant build.
+Ant has no layout convention, so Supercov reads the build: it follows the
+`<import>` chain with the properties it defines, takes the directories each
+`<batchtest>` or `<testclasses>` scans as the tests — or, where it scans
+compiled classes, the sources the `<javac>` writing them compiles — and what
+its other `<javac>` tasks compile as the code they test, leaving out the
+directories the build writes into. Java elsewhere, such as the sources a test
+compiles as its input, is left alone. Where a `<javac>` compiles from a path
+the build makes only at run time, every other `src` directory holding Java is
+taken as product code too.
+When the build names its tests where that cannot be read — a path a custom
+task builds at run time — say where they are:
+
+```sh
+SUPERCOV_TEST_ROOTS=test/src npx supercov -- ant test
+```
+
+`SUPERCOV_SOURCE_ROOTS` narrows the product code the same way as for any
+project.
+
+Ant's `<junit>` task runs JUnit 4 itself, not on the JUnit Platform, so in the
+copy each `<junit>` task gets a result formatter of Supercov's next to its own
+formatters. The task loads it in every JVM it forks — one per test class by
+default — and it tells the runtime when each test starts and ends. Each JVM
+writes its evidence at the end of its suite and on shutdown, so a JVM the
+task stops for running past its timeout still leaves what it recorded, and
+the run merges every one. The formatter is compiled by the copy's build just
+before the task, against the task's own classpath and Ant's runtime. Tests
+keep the names Ant's XML reports give them, `Class#method`; a test a suite
+runs is named within it, `Suite#Class#method`, and a test two tasks or
+filesets both run is numbered in the second (`[2]`). A test the task stops
+for its timeout is recorded as failed, with what it ran until then. A build
+that names no tests where they can be read keeps the `src/main` and
+`src/test` convention.
+
+Ant's `<junitlauncher>` runs tests on the JUnit Platform, so it gets the same
+platform listener Maven and Gradle builds get, as a `<listener>` of the task,
+compiled the same way against the task's classpath. TestNG's `<testng>` task
+gets the TestNG listener in its own `listeners` attribute, beside any the
+build names.
+
+Many Ant builds import a shared harness from beside the project —
+`<import file="../../../tools/master/build/common/common-build.xml"/>` — and
+the `<junit>` task lives there. The copy is laid out at the project's place
+within the closest directory holding it and everything its build imports,
+with links standing for the rest, so `../` reaches what it reaches in the
+checkout. A file outside the project that holds a `<junit>` task — the
+harness, or a wrapper script's own build file beside it — is changed in the
+copy's own version of that file; the original and the rest of the harness
+are only read. Small files and directories beside the project are copied
+rather than linked, so a wrapper script that writes beside itself writes into
+the copy; anything larger than 4 MB stays a link.
+
+Only the project's own sources are measured: modules resolved as jars, Ivy's
+included, are not instrumented. Code that is not Java or Kotlin — XSLT,
+sitemaps, scripts run by the JVM — is outside what the figure covers.
+
+```sh
+npx supercov -- ant test
+npx supercov -- ../../../run-single-test.sh . ALL
+```
+
 ## Containers, VMs, and remote execution
 
 Supercov can collect from supported processes launched through a container, VM,

package/docs/workspace-isolation.md

@@ -200,8 +200,9 @@ its settings with paths moved to the guest's mount, and the guest's coverage is
 credited to the test that launched it. A guest started any other way, with none
 of those settings, still runs the instrumented code: it reads the run's
 settings from the mounted workspace itself, and its coverage counts for the run
-as background coverage, credited to no test. The run summary's `Guests` line
-says how many processes did that.
+as background coverage, credited to no test. The run summary's `Unattributed`
+line counts processes whose coverage has no named test identity. Local processes
+can also appear here; this count does not establish container or VM execution.
 
 Dependencies stay out of the instrumented copy. The root `node_modules` is
 linked entry by entry to the project's own, so an environment that mounts the

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "supercov",
-  "version": "4.22.0",
+  "version": "4.23.0",
   "description": "Coverage, security and code quality for coding agents",
   "keywords": [
     "coverage",
@@ -122,14 +122,14 @@
     "test:windows-tls": "node scripts/windows-tls-test.mjs"
   },
   "optionalDependencies": {
-    "@supercov/cli-darwin-arm64": "4.22.0",
-    "@supercov/cli-darwin-x64": "4.22.0",
-    "@supercov/cli-linux-arm64-gnu": "4.22.0",
-    "@supercov/cli-linux-arm64-musl": "4.22.0",
-    "@supercov/cli-linux-x64-gnu": "4.22.0",
-    "@supercov/cli-linux-x64-musl": "4.22.0",
-    "@supercov/cli-win32-arm64": "4.22.0",
-    "@supercov/cli-win32-x64": "4.22.0"
+    "@supercov/cli-darwin-arm64": "4.23.0",
+    "@supercov/cli-darwin-x64": "4.23.0",
+    "@supercov/cli-linux-arm64-gnu": "4.23.0",
+    "@supercov/cli-linux-arm64-musl": "4.23.0",
+    "@supercov/cli-linux-x64-gnu": "4.23.0",
+    "@supercov/cli-linux-x64-musl": "4.23.0",
+    "@supercov/cli-win32-arm64": "4.23.0",
+    "@supercov/cli-win32-x64": "4.23.0"
   },
   "peerDependencies": {
     "@playwright/test": ">=1.55.0",