PARASOFT / ESL
DEMONSTRATION PLAN

Apollo / Autonomous driving / Customer demonstration proposal

A real automotive codebase.
A measurable quality journey.

Apollo is Baidu's open-source autonomous-driving platform: perception, localization, prediction, planning, and control, connected through Cyber RT middleware.

LOCAL C / C++ SOURCE INVENTORY471,310

lines of code, excluding comments and blank lines

4,313C / C++ source & header files
449,643code lines excluding third_party

C++: 447,579 code lines · C: 23,731 code lines
703,440 physical lines, including blanks and comments.

Languages & roles

C++ / C for native components; Python / Shell for tools and automation; JavaScript / TypeScript for Dreamview UI; Protocol Buffers for message schemas. This LOC count covers C/C++ only.

Build & analysis setup

Linux + Docker + Bazel. Capture compiler invocations with Parasoft cpptesttrace, then analyze the agreed targets. The prior local scan used //cyber/..., not the entire repository. GPU dependencies vary with the selected modules.

Sense & localizePredict & planControl the vehicleTest & improve
Readiness gate: this checkout defaults to C++14 in tools/bazel.rc; a C++17 configuration also exists. Confirm effective compiler flags and language-mode alignment for MISRA C++:2023 before the baseline. Agree a separate ruleset for C sources.

Inventory: local checkout, 04 October 2026, scc 4.1.0; C/C++ source and header extensions; repository ignores respected; .git and .cache excluded. Repository inventory is not scan coverage.
Sources: Apollo README, tools/bazel.rc, local LOC measurement · https://github.com/zuwasi/apollo

← Swipe to navigate →   ·   Arrow keys / Space   ·   Next: the animated demonstration workflow

Customer review / Demonstration plan / 04 October 2026

Parallel analysis.
Verified improvement.

Six Apollo pull requests. One controlled cycle from MISRA findings
to AI-assisted fixes, tested code, and visible results.

6PRs / SERVERS
1DTP SERVER
100FINDINGS TO FIX
01 / ANALYZE IN PARALLEL · SELECT ANY COMPONENT TO EXPLORE
Parasoft C/C++testMISRA C++:2023
↗ Incredibuild accelerationShared coordinator + helpers
Each server analyzes its PR's full agreed scope, not one-sixth of the code.IB is planned; the same demo can run without acceleration.
02 / REMEDIATE, TEST & MEASURE
ACCEPTANCE TARGETS100 selected findings resolvedNo new findingsTests passSame analysis scope
01
GUIDED WALKTHROUGH · NOT LIVE EXECUTION

Freeze six independent PR revisions

Apollo supplies six exact PR revisions. Agree the build targets, compiler mode, and exclusions so every before/after comparison uses the same scope.

01 / Parallel execution

Six independent PRs.
One consistent process.

Jenkins assigns each PR to a Parasoft automation server. Each run has its own checkout, build capture, report identity, and quality gates.

↗ Incredibuild accelerationShared coordinator + helpers
Concurrency increases throughput across PRs. Incredibuild accelerates the supported workload. The demonstration measures actual timings; no speedup factor is promised.
PLANNED PIPELINE

PR 01 → Server 01

InputExact PR revision
AnalysisMISRA C++:2023
ScopeComplete agreed build targets
DTP historyPR 01 / before & after
Run identityPR + commit + Jenkins run

All six publish to one DTP server. PR-specific views keep branch results separate; shared findings are not counted six times as unique defects.

Readiness check: six concurrent analysis licenses, unit-test entitlements, suitable server capacity, and an agreed Apollo build scope. If IB is unavailable, continue on the six Parasoft servers without it.

02 / The improvement loop

Fix the finding.
Keep the behavior.

Amp assists with remediation. Builds, tests, static analysis, and human review determine whether a change is accepted.

Animated explanation, not scan results. Select a phase to read it at your own pace, or use the motion control above.

If a build, test, or scan gate fails, return to the fix stage. Do not accept or merge the change merely because the finding count is lower.
TARGET, NOT A MEASURED RESULT

Select 100 distinct findings

Choose 100 actionable findings total across the six PRs. Record each rule ID, source location, baseline revision, and assigned PR. Prefer non-overlapping files and confirm the test baseline before editing.

100selected findings
0selected findings remaining

This is the target for the selected set, not a claim that Apollo has only 100 violations or will become violation-free.

Also requiredNo new findings + passing tests
Comparison controlSame rules + analysis scope

03 / Customer decision

Agree the scope.
Then run the demonstration.

Review these four items together. Download the review summary and return it through your normal approval process.

PROPOSAL / AWAITING CUSTOMER APPROVAL

The deliverable is evidence.

6 / 6 / 1 PRs / servers / DTP

Per-PR baseline and final reports, reviewed patches, test results, actual timings, and a traceable summary of the 100 selected findings.

Current presentation modeWith Incredibuild
Items reviewed0 of 4

Review selections are local to this page.

This page does not submit or record customer approval. No personal information is collected. A successful demo is not a claim of full MISRA compliance or production safety certification.

Project: https://github.com/zuwasi/apollo · Presented by ESL, Engineering Software Lab.

WORKFLOW COMPONENT