How to fuck up software releases

A developer shares a series of hard-learned mistakes made while managing software releases for open-source projects, and how each mistake led to an incremental improvement in their release automation script. Issues covered include wrong version numbers, bad changelogs, version files not committed, scripts not checked into version control, Busybox vs GNU tool differences causing tarball mismatches, releasing from a feature branch, and tagging without pulling first. Each problem prompted a fix to the semver script, illustrating how release automation evolves through accumulated failures.