Topic · test impact analysis

Test impact analysis

Gist

Test impact analysis is which tests to rerun for this change. Proof runs proof test affected. Azure TIA reads a call-graph. Proof reads requirement traces and Verifies annotations. Incomplete evidence falls back to the full suite. Jama still authors.

proof test affected

Keep Azure if you already run coverage TIA. Keep Datadog if you already map tests to files. Proof will not treat a guessed subset as the plan.

Request demo Last full suite vs this plan

01 · The missed slice

The suite is green. The change never named its tests.

CI reran everything. The shall that moved had no Verifies row. The focused plan never existed.

Martin Fowler and Azure will tell you test impact analysis is a call-graph: which procedures this diff reached, which tests cover those procedures, run those tests. Datadog stores coverage per test and intersects it with the files in the PR. That is not this URL. Proof's proof test affected is the conservative plan from authored-delta requirements, changed files, implementation links, hierarchy, and Verifies test annotations.

The motivating miss is mechanical. SYS-REQ-010 changed. The full suite ran for twenty minutes and passed. Nobody asked which tests were tied to that shall. The tests that would have failed were not in the suite that night, or they were, and nobody can show they were selected on purpose. A green full run is not a named plan.

The command does not build a call-graph. It does not train on your build history. If ownership or evidence is incomplete, it falls back to the configured full test command and prints the reason. The blast radius of the shall stays on software change impact analysis. Duration stays on slow tests.

proof test affected
proof test affected --run

The first command is the plan. The second runs it. A green suite next to an unnamed shall is how the unselected test stayed dark.

02 · The exhibit

Same diff. A full suite, or this plan.

The shall moved. The tests were not named. Click the tabs.

The run

Command the configured full suite
Exit 0
Why everything that was wired ran

Suite green

This hop

No list of tests tied to the shall that moved. The selection is a guess.

No plan

The run

Keep the full suite. Use --full when you need to prove it ran. That is not the default hop.

Keep the suite

Proof

Ask authored delta, files, Verifies rows
Out focused commands, or the full fallback with a reason

A green suite is not a named plan

Same diff. A full suite, or this plan. Click the tabs.

Surface	What they do	What Proof does	What we lose
Azure / Datadog TIA	A coverage call-graph. Tests that reached the changed procedures.	Requirement traces and `Verifies` annotations.	Not the call-graph. Keep the coverage map they already run.
CloudBees predictive selection	A model trained on build history.	A deterministic plan from authored evidence.	Not the model. No history score.
Software change impact analysis	The blast radius of the shall. software change impact analysis.	The test plan for that delta.	Not `proof trace impact`.
Jama	The authoring programme. Tests live on the item if you put them there.	Fail the missing plan on the files. Fallback reason in the log.	Not Jama's test matrix. Jama still authors.

The teaching graph is still a changed shall next to an unnamed suite. Run the plan before you trust the green. Three traps ship through hand-edits that did not name the tests: full-suite-only, focused commands on incomplete evidence, and a guessed subset with no fallback reason.

proof test affected
proof test affected --base origin/main
proof test affected --requirement SYS-REQ-010 --format json
proof test affected --run
proof test affected --full --run

JSON is for the agent. --run executes the plan. --full is the proof that the whole configured suite ran. Incomplete traces print the reason and fall back. See software change impact analysis if the failure is the shall's blast radius, not the test list. See slow tests if the failure is duration, not selection.

03 · The honest loss

Proof names the plan. It does not build the call-graph.

A focused run is not a proof of the Go. Jama still authors.

Proof does not implement Azure Test Impact Analysis, Datadog's coverage map, or CloudBees predictive selection. It does not walk a call-graph. It does not train on prior builds. If a requirement has no supported Verifies evidence, the hop is the full suite, not a guessed subset. A green focused run is not proof the Go matches the shall. We have not scored this against a frozen Azure TIA corpus. The loss is named, not scored.

Blast radius stays on software change impact analysis. Duration stays on slow tests. Jama still authors.

Request demo How the install is bought

04 · Nearby questions

What people type next.

What is test impact analysis? Same question. Same URL.

What is regression test selection? Same cluster. The academic name for picking a subset after a change. This URL is the plan Proof can emit.

What is test selection? Hiring tests and diagnostic kits. Not this URL.

What is software change impact analysis? The blast radius of the shall, not the test list. software change impact analysis.

What are slow tests? Duration of a named leaf, not which tests to run. slow tests.

Is Proof a Jama alternative for test matrices? No. Jama still authors. Proof vs Jama.