shio-build-test
Skill3 starsChanged 8 days agoApache-2.0View on GitHub
Read before building, running or measuring Shio's test suite, or when diagnosing a build/test oddity. Triggers on mvnw, skip.npm, frontend-maven-plugin, surefire/failsafe, "Nothing to compile", counting tests, the pnpm workspace, Vite config, publishing to npm, argLine and the Mockito javaagent, benchmark isolation, and any failure that looks unrelated to the change you just made.
In this repository14 other files it gives its agents.
- agents.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- roadkeepSkill
- shio-agent-surfaceSkill
- shio-console-reviewSkill
- shio-console-uiSkill
- shio-content-apiSkill
- shio-content-filesSkill
- shio-databaseSkill
- shio-docs-syncSkill
- shio-docs-writeSkill
- shio-rendererSkill
- shio-replication-testSkill
- shio-roadmap-docsSkill
The file
Files
Building, testing and measuring Shio
Backend-only work: always clean, always -Dskip.npm=true, always the full suite.
./mvnw -pl shio-app -am clean test -Dskip.npm=trueCount the tests from the run, not from target/:
./mvnw -pl shio-app -am clean test -Dskip.npm=true 2>&1 | grep -E "Tests run:.*Failures.*Errors" | tail -1Read docs/agents/build.md for the Maven/pnpm wiring, Vite config, static-resource mapping and the npm publish order, and docs/agents/troubleshooting.md for the catalogue of known-weird failures.
Four artefacts of an earlier command changing what a later one says
This is the class of bug to suspect first when a result makes no sense:
cleanis not optional (SH282). Without it, a call to a method that does not exist prints "Nothing to compile — all classes are up to date" and BUILD SUCCESS: the IDE compiles intotarget/classeson save, so the.classoutlives the.java. Eclipse JDT emits a class withjava.lang.Error: Unresolved compilation problemembedded, so it fails at runtime in whatever test calls it — one missing accessor read as 53 errors across five unrelated classes.-Dmaven.compiler.useIncrementalCompilation=falsedoes not help. Treat "Nothing to compile" as no information; the tell is aServletExceptionwrappingjava.lang.Error.-qhides the total, andtarget/surefire-reports/accumulates (SH283).-qsuppresses surefire'sResults:line — the only authoritative count — and summing the reports double-counts any class you ran earlier with-Dtest=X(977 claimed vs 974 real).-Dskip.installnodepnpm=true -Dskip.pnpm=truedo nothing here. Those arefrontend-maven-plugin's per-goal properties; this pom is wired for the single<skip>${skip.npm}</skip>toggle, so passing them runs the React build anyway.- A benchmark that measures the whole instance must not share a datasource
(SH315).
ShBenchmarkIsolationLintTestenforces three source rules — the convention was documented on four classes and broken five times, always caught by accident, because the code is right and only the measurement is wrong.
Mockito is a -javaagent, not a self-attach (SH250)
Surefire and failsafe carry @{argLine} -javaagent:${org.mockito:mockito-core:jar}.
@{argLine}, never ${argLine} — JaCoCo sets that property, so the plain form
replaces its agent and coverage silently stops. If you add another agent, append the
same way. Self-attaching made ShTenantSuspendedAccessTest fail six of six once and
pass on every re-run; a flake is worse than a break, because it teaches people that
red might mean nothing.
Before you quote a benchmark number
Know which way it moved and why — Blocks N and O made the baseline lean rather than
the agent path fast, so the ratios fell three blocks running with the agent path
untouched. Quote a source, not a task ratio; the aggregate ≈10× is retired.
Ceilings live in shio-app/src/test/resources/token-budgets.properties; raising one
is allowed and must be justified in the commit.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

