Every requirement satisfied, still broken: dependency bugs I found in six Python apps
pip check passing doesn't mean your packages work together. It means each package's declared version...
Tag archive
pip check passing doesn't mean your packages work together. It means each package's declared version...

This is part 7 of the series "Protect Your Own Service Yourself — Self-Defense Security in the Age of...
The Anatomy of Bloat: Tracing the Legacy Dependency Tree When investigating why test...
Introduction: The Dependency Dilemma Managing software dependencies is a high-wire act. On...
Introduction: Pip 26.3 Release and the Persistent Security Tightrope in Python...
The suggested API is real. The documentation example is valid. The project still cannot use it. I...
An agent lands a small bug fix but regenerates the lockfile from a clean install, pulling in dozens of unrelated transitive version bumps that ride inside the same diff. Covers why lockfile churn hides supply-chain changes, how to pin a single package instead of reinstalling, and why dependency bumps belong in their own reviewed PR.

Build a dependency inventory, decide what to trim or rebuild, fuzz the code you own, and check compliance on each change.

I installed 13 popular packages and got 389 packages and 59.7 MB in node_modules. Then I replaced them with what ships in Node: --watch, --env-file, fetch, node:test, type stripping and more. Every swap has a test, and the gotchas are listed, including a flag Node 26 removed.
Why unit tests stay green through a major dependency bump, the three quiet breaks mocks cannot see, and the cheap extra pass that catches renamed fields, dropped options, and changed callbacks before merge.
What lock files and hash pinning actually guarantee, how signature and provenance verification differ from it, and where the process fails in practice.
Spent a day on a CI build that worked locally and failed on the runner. Same Node version, same lock...