Release 1.0.1 - #10
Merged
Merged
Conversation
Two bug fixes and an internal search optimization have landed since v1.0.0. No public API was added, changed, or removed, so this is a patch release under semver. Make LINEAR_SCAN_LIMIT a private constant. It is a tuning value used only by Node#key_index, but Ruby constants are public by default, so leaving it exposed would have made this a minor release and locked the value into the public API. Doing it now avoids a breaking change later. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
travis-ci.org shut down in 2021 and the config still pinned rvm 2.4.1, so nothing has been running it. Its deploy block was also the only automated path to RubyGems; releases are now manual via `rake gem:release`. Drop the now-dead .travis.yml entry from the Bones exclude list as well. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The tests use the context/should DSL, which comes from shoulda-context. The shoulda meta-gem also pulls in shoulda-matchers, whose Rails model assertions the suite never uses, and which depends on activesupport. Depending on shoulda-context directly takes the lockfile from 30 gems to 14, dropping activesupport, i18n, tzinfo, concurrent-ruby, connection_pool and the rest of that tree. shoulda-matchers has been pinned at 3.1.3 (2016) by shoulda's ~> 3.0 constraint, and was emitting method-redefinition warnings on every run. Those are now gone. This does not change the supported Ruby range: 3.2 remains the floor, now set by minitest 6, erb 6 and bundler 4 rather than by activesupport 8.1. These are all development dependencies, so the published gem is unaffected either way -- it still declares no runtime dependencies. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replaces the Travis config removed in the previous commit. Runs the suite on Ruby 3.2, 3.3 and 3.4 for pushes to master and for pull requests. Exclude .github from the gem manifest, for the same reason .gitignore and .ruby-version are excluded. Also exclude ^pkg/. Setting exclude at all replaces the Bones defaults, one of which was ^pkg/, so build products from previous releases sitting in the working tree were being swept into the manifest -- the 1.0.0 gem, its tarball and its unpacked tree were all listed in s.files, and two pkg/ paths had ended up in extra_rdoc_files. Releasing from a tree with a populated pkg/ would have shipped the old gem inside the new one. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Everything else already resolved to its latest release. The one exception is rdoc, held at 6.17.0 by bones' `rdoc (~> 6.0)` pin -- 8.0.0 is out but is not reachable while bones is the build tool. The suite passes unchanged on shoulda-context 2.0, and the major bump clears the last "assigned but unused variable" warning from the test run. It declares no required_ruby_version, so the 3.2 floor and the CI matrix are unaffected. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bones 3.9.0 is the end of its line: it still requires net/smtp, which left the stdlib in Ruby 3.1, so every rake invocation opened with load errors. It also pinned rdoc to ~> 6.0, holding the project two majors back, and its exclude list was what let build products under pkg/ leak into the gem manifest. btree.gemspec is now checked in and hand-written. It reads version.txt so that stays the single source of truth for both the package and Btree.version, and it builds its file list from git ls-files, which makes the pkg/ leak structurally impossible rather than a matter of remembering an exclude entry. It also declares the MIT license the README always claimed -- Bones was shipping licenses = [nil] -- and required_ruby_version >= 2.3.0, the floor set by Array#bsearch_index in Node#key_index. The Rakefile gains explicit publishing tasks. release:check refuses to proceed on a dirty tree, an existing tag, a version already on rubygems.org, or a missing History.txt entry, and warns when the branch has drifted from origin/master. release:tag and release:push are separately invokable for recovery when one half of a release fails. The full release task runs the checks and the suite before either. release:push depends on repackage rather than build. The packaged gem is a file task, so an existing pkg/*.gem newer than its sources satisfies build without rebuilding, which is harmless locally and wrong when the next step is an irreversible push. Dropping bones takes the lockfile from 14 gems to 6. The Gemfile now uses the gemspec directive, and CI moves from rake test:run to rake test. Verified by installing the built gem into a throwaway GEM_HOME: Btree.version resolves version.txt from the installed layout and queries return correctly. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Ordering: everything reversible now happens before anything that is not. A tag is cheap to delete, so it goes first; the gem push is permanent -- RubyGems refuses to re-push a version even after a yank -- so it goes last, and only after the packaged gem has been built and exercised. An earlier note in review had this backwards. release:check now reports every problem at once instead of aborting on the first, and gained two. It requires HEAD to exist on some remote branch, so a published gem can never be built from work that only exists on one laptop. And it checks for rubygems credentials up front, because the alternative is discovering a missing or wrongly-scoped key as a 403 after the tag is already pushed. Branch drift stays a warning rather than an abort, so releasing from a release branch is still possible. release:verify is new: it rebuilds the gem, prints the manifest, installs it into a throwaway GEM_HOME and runs a smoke test against the installed copy with Bundler's environment stripped. This is the only check that covers what was actually packaged rather than what happens to sit in the working tree -- a file missing from the manifest looks perfectly fine locally and only breaks for whoever installs it. release:push now states plainly that the push is irreversible and requires the version number to be typed back before proceeding. In a non-interactive shell it refuses outright unless FORCE=1, rather than hanging on a prompt nobody can answer. release:confirm checks afterwards that the version really is live. The gemspec declares rubygems_mfa_required, so a leaked API key is no longer sufficient to publish on its own. This requires MFA enabled on the rubygems.org account or the push is rejected. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
release:check hardcoded ~/.gem/credentials, but modern RubyGems stores the key at the XDG path (~/.local/share/gem/credentials), so a signed-in machine was reported as having no credentials. Ask Gem.configuration.credentials_path instead of guessing. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
v4 targets Node 20, which GitHub is deprecating and now forces onto Node 24, producing a warning annotation on every job. v5 runs on Node 24 natively. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Prepares the 1.0.1 patch release and modernizes the build/release tooling.
Release contents (1.0.1)
Two bug fixes plus an internal search optimization, all backwards-compatible — semver PATCH:
LINEAR_SCAN_LIMITmade a private constant so it stays out of the public API (keeps this a clean patch bump).Tooling
net/smtp(gone from stdlib in 3.1) and pinned rdoc two majors back.release:checkaggregates pre-flight checks (clean tree, HEAD on a remote, no existing tag/version, History.txt entry, credentials present);release:verifyinstalls the packaged gem into a throwaway GEM_HOME and smoke-tests it;release:pushis irreversible-by-design with a typed confirmation and tag-before-push ordering. Gemspec declaresrubygems_mfa_required.Verification
rake testgreen throughout (17 runs, 92 assertions). The built gem was installed into an isolated GEM_HOME and exercised (Btree.version, range queries, lookups). This PR is also the first run of the new CI workflow.🤖 Generated with Claude Code