Ruby .gitignore template

The leading slash is the defining feature of this template. A pattern such as /tmp/ matches only a tmp directory at the repository root, while tmp/ without the slash would match one at any depth. Ruby projects have a lot of common directory names, so the anchoring matters: it keeps /coverage/ from swallowing a coverage directory inside a fixture, and /pkg/ from matching an unrelated package folder deeper in the tree.

Gemfile.lock is the judgement call, and this template makes the choice appropriate to a gem rather than an application. A gem is a library: its lockfile pins only what its own CI tested against, and consumers resolve their own versions, so tracking it adds noise without adding safety. An application is the opposite case. Deploying without a committed lockfile means production can resolve a different set of gems than you tested, so if this is an app, delete the line.

The bundler rules cover /.bundle, which stores the local install path configuration, and /vendor/bundle, which is where gems land when you install them into the project rather than system wide. Both are the result of running bundle install and neither should travel. /lib/bundler/man is generated documentation from the same tool.

*.gem and *.rbc are the packaged gem file and compiled Ruby bytecode. The test output rules cover /spec/reports, /spec/examples.txt and /test/tmp, all of which are regenerated on every run and produce conflicts if two people commit them from different machines.

.byebug_history is worth calling out. It is the debugger's command history, and it records every expression you evaluated at a breakpoint. On a session spent debugging an authentication problem, that history very plausibly contains a token or a password you printed to check it.