software
Witness
A real Selenium WebDriver/pytest E2E suite (Page Object Model, RBAC-UI and forced-browsing tests) against Rollkeeper, plus real Locust load results -- rebuilt from an original Selenium IDE report.
25/25
e2e: Selenium tests passing (real browser, live Rollkeeper app)
0
e2e: Ruff lint issues
6
Real bugs found and fixed (5 in e2e/, 1 in Rollkeeper itself)
0.0% -> 88.7%
Locust load test: failure rate at 1 vs. 20 concurrent users
Originally a non-functional (load/stress) testing exercise using Selenium IDE's record-and-playback tooling against two targets: a university's student-portal login page and Wikipedia's search engine. See the Ethical Notice regarding testing a real, live third-party system without documented authorization. Rebuilt with e2e/: a real, version-controlled Selenium WebDriver 4.48 + pytest suite (Page Object Model, Page Component Objects, explicit waits, Selenium Manager) retargeted entirely at Rollkeeper (a local app this same effort built) -- login, CRUD, search, pagination, and real browser-level RBAC/forced-browsing verification, complementing Rollkeeper's own server-side pytest suite. Building it found and fixed six real bugs, including a genuine rate-limiting bug in Project #7 itself. A separate Locust HTTP-level load experiment (1/5/10/20 simulated users) produced real, committed quantitative results.