Skip to content

Import codebase from private repo and prepare 0.1.0 - #1

Merged
joe-sharp merged 14 commits into
mainfrom
import-and-release-0.1.0
Nov 8, 2026
Merged

joe-sharp merged 14 commits into
mainfrom
import-and-release-0.1.0

Conversation

@joe-sharp

@joe-sharp joe-sharp commented Nov 8, 2026 •

Copy link
Copy Markdown
Owner

Summary

  • Imports the codebase from the private repo
  • Turns the e2e and appraisal-e2e CI jobs back on
  • Uses an absolute URL for the README banner so it renders on rubygems.org
  • Dates the 0.1.0 entry in the CHANGELOG

Test plan

  • CI passes on this branch (including e2e and appraisal-e2e jobs)

🤖 Generated with Claude Code

@joe-sharp joe-sharp self-assigned this Nov 8, 2026
@joe-sharp
joe-sharp force-pushed the import-and-release-0.1.0 branch from 95a8ecf to 02292b2 Compare November 8, 2026 18:18
Copies bundle_matrix-private at 7f54213 (origin/main), then:

- Turn the e2e and appraisal-e2e CI jobs back on
- Use an absolute URL for the README banner, so it shows on rubygems.org
- Ship LICENSE.txt in the gem, matching the private repo
- Date the 0.1.0 entry in the CHANGELOG

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@joe-sharp
joe-sharp force-pushed the import-and-release-0.1.0 branch from 02292b2 to 56bb5cd Compare November 8, 2026 18:20
joe-sharp and others added 13 commits November 8, 2026 12:29
narrow_update read the secondary gemfile's declared groups on every
update, which evaluates the whole gemfile. A gemfile that raises when
evaluated then failed a plain 'bundle update <gem>' that never needed
its groups.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Bundler's Thor keeps the last occurrence of a repeated switch, but
option and option? used the first. The plan then judged a flag by
a value Bundler never used, and replace_values forwarded the wrong
--group or --path to every child.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
--source takes gem names, but its values were forwarded unfiltered,
and unmatched? ignored them. A gemfile without the named gem ran
'bundle update --source' and failed, and one with it was skipped
when only --source named it.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Gems that came after a bare -- were merged into the gem list, and
to_argv printed them before the options with no terminator. A gem
such as -weird was then read by the child as an unknown switch.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The gemspec pinned bundler to the tested range, so a user on a newer
Bundler could not install the plugin, and the untested-Bundler warning
could never run. The gemspec now requires MIN_BUNDLER as a lower bound,
and TESTED_BUNDLER only drives the warning.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Invocation.parse relies on Thor internals and ran before the
matrix.disabled and BUNDLE_MATRIX_ACTIVE checks. A parse error then
broke every bundle install, even with the plugin disabled and in the
child runs.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Only add --path was rewritten. A relative --binstubs or
--target-rbconfig was forwarded as typed, and the child resolved it
from the gemfile's own directory instead of the project root. Install
binstubs and target-rbconfig now get the same rewrite.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The note claimed --path is not forwarded, but Bundler 2 saves it to
.bundle/config, and ProjectSettings passes that on as BUNDLE_PATH to
every child. The note now says that, which is also accurate on Bundler
4, where --path is not saved.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Discovery removed the main gemfile by comparing expanded path strings.
A symlink to it, or the same file reached another way, stayed in the
list, so the main Gemfile was installed or updated twice.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The tag was pushed before the gem was built. A build failure, such as
an invalid gemspec, left a public tag for a version that was never
released, and the source_code_uri metadata pointed at it. The build now
runs first, and the tag is only created and pushed after it succeeds.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Shorten one example name and split the disabled and secondary-run
parse checks into two examples, so each stays within the style limits.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Steep reported the new Planner, Discovery and version members as
undeclared, and still listed the removed rewrite_path. The signatures
now match the code.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The tag was pushed before the gem was published, so a failed publish
left a public tag for a version that was not released. The release now
builds, creates the tag locally, pushes the gem to RubyGems and gem.coop,
and pushes the tag only at the end. A rerun reuses the local tag and skips
versions already published.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
@joe-sharp
joe-sharp merged commit 6dd744e into main Nov 8, 2026
56 checks passed
@joe-sharp
joe-sharp deleted the import-and-release-0.1.0 branch November 8, 2026 19:03
@joe-sharp
joe-sharp restored the import-and-release-0.1.0 branch November 8, 2026 19:15
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant