| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
| |
Call through to the stub instead of materializing the full remote_spec since all gems added to the local index are asked whether they are missing extensions or not
|
| |\
| |
| |
| |
| |
| |
| | |
Fail gracefully when installing a spec where the API is missing deps
Fixes https://github.com/bundler/bundler/issues/5426
Closes https://github.com/bundler/bundler/issues/5339
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| |\ \
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
[Resolver] Improve sorting of deps with 0 or 1 possibility
As the added comment says, this will force all dependencies where there are only 0 or 1 total specs with that name to be resolved first. Right now, they sort ~last, which leads to poor performance.
(This was one way of fixing some recent resolver issues, even though we've done the proper fixes in Molinillo for them)
|
| | | | |
|
| | |/ |
|
| |\ \
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Enable compact index when OpenSSL FIPS mode is enabled but not active
Fixes #5433. Since there is no easy accessor in Ruby to detect whether or not FIPS mode is currently active, the best approach I could come up with is to `fork` a separate process and attempt to generate a build MD5 object as a test of whether MD5 module is currently available.
Because `fork` approach won't work on some platforms (JRuby, Windows etc), `md5_supported?` returns `false` on any platforms where FIPS mode is enabled and `Process.respond_to?(:fork)` is `false`.
I've added a spec that simulates behavior when OpenSSL FIPS mode is active - an error message is output to STDERR and the process is killed with the `ABRT` signal.
|
| | | | |
|
| | | | |
|
| | | | |
|
| |\ \ \
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Add documentation for mirror.fallback_timeout option.
As suggested in https://github.com/bundler/bundler/pull/5403#issuecomment-278384552 instead of changing parsing `mirror.fallback_timeout` I added documentation for it.
Pull request https://github.com/bundler/bundler/pull/5403 can be closed.
|
| | | | | |
|
| | | | | |
|
| |\ \ \ \
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Fix Github username for README.md contrib URL
This fixes #5438 where the wrong git config variable was being
retrieved.
This was using `git config user.name` rather than `git config github.user` in URL in `README.md`
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
This fixes #5438 where the wrong git config variable was being
retrieved.
Renamed user name to author name & specs added
`git_user_name` renamed to `git_author_name` and tests added for genuine
`git_user_name` to verify proper URL in README.md.
Renamed `git_user_name` to `github_username` per @segiddins request.
Rubocop style fix on hash k,v spacing
Replace backticks with `sys_exec` per @segiddins. Hat-tip to
@olleolleolle for explaining `sys_exec` from `spec_helpers/support.rb`
is a helper method that smooths the path for the RSpec usage in the
project. (Thank you.)
|
| |\ \ \ \ \
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
fixing small typo in comments
Title says it all, really.
|
| |/ / / / / |
|
| |\ \ \ \ \
| |_|_|/ /
|/| | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
improvements to the BUG_TRIAGE guide
\cc @indirect What we were discussing earlier
This PR consists of several changes like:
* Update references to guides (troubleshooting.md & issues.md specifically)
* Adding references to existing guides (pull_request.md)
* General cleanup, adding clarity etc
|
| | | | | | |
|
| | |/ / / |
|
| |\ \ \ \
| |/ / /
|/| | |
| | | |
| | | |
| | | |
| | | |
| | | | |
remove post-install on account of confusion
as demonstrated on twitter[1], this message was unfortunately convincing people that Ruby Together pays for the rubygems.org servers. let's remove it and come up with something clearer later.
[1]: https://twitter.com/MelissaKaulfuss/status/829504864229023744
|
| |/ / /
| | |
| | |
| | |
| | |
| | | |
as demonstrated on twitter[1], this message was unfortunately convincing people that Ruby Together pays for the rubygems.org servers. let's remove it and come up with something clearer later.
[1]: https://twitter.com/MelissaKaulfuss/status/829504864229023744
|
| |\ \ \
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
r=segiddins
Defer requiring rubygems/spec_fetcher until it becomes necessary
Avoid conflict between two versions of openssl gem on 'bundle exec' if
a newer and non-default version of openssl gem is installed to the
system. rubygems/spec_fetcher loads openssl through resolv and
securerandom when running with Ruby <= 2.4.
This is not a proper fix for #5235 as other standard libraries that
Bundler currently loads will be gemified as well in Ruby >= 2.5,
however, this will fix openssl's case.
Reference: https://github.com/bundler/bundler/issues/5235
Fixes: https://github.com/ruby/openssl/issues/103
---
Reproduce:
1. Install Ruby 2.4.0 (comes with openssl 2.0.2)
2. Run `gem install openssl` to install openssl 2.0.3
3. Create a Gemfile and run `bundle install`
4. Run `bundle exec ruby -e'require "openssl"'
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Avoid conflict between two versions of openssl gem on 'bundle exec' if
a newer and non-default version of openssl gem is installed to the
system. rubygems/spec_fetcher loads openssl through resolv and
securerandom when running with Ruby <= 2.4.
This is not a proper fix for #5235 as other standard libraries that
Bundler currently loads will be gemified as well in Ruby >= 2.5,
however, this will fix openssl's case.
Reference: https://github.com/bundler/bundler/issues/5235
Fixes: https://github.com/ruby/openssl/issues/103
|
| |\ \ \ \
| |_|_|/
|/| | |
| | | |
| | | |
| | | |
| | | | |
Don't read global settings on a read-only FS with no $HOME
Closes https://github.com/bundler/bundler/issues/5371
This is an alternative to https://github.com/bundler/bundler/pull/5385
|
| | | | |
| | | |
| | | |
| | | | |
and $TMPDIR is not writable
|
| | | | |
| | | |
| | | |
| | | |
| | | | |
This was, modifications to the temporary directory are blown away
between each test and don't clutter the user's FS
|
| |\ \ \ \
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
[Thor] Update to v0.19.4
Closes #4876
@rafaelfranca
|
| | | | | | |
|
| | | | | | |
|
| |\ \ \ \ \
| |/ / / /
|/| | | |
| | | | | |
Version 1.14.4
|
| | | | | | |
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
[Fix #5358] Handle files conflicts when using `bundle gem`
`bundle gem` command fails with an exception when creating a gem having
the same name of a file in the specified path.
For instance:
> touch hello
> bundle gem hello
This bug is due to Thor gem. At the moment, Thor does not handle
conflicts between files and directories. This commit mitigates the
problem by rescuing the exception from Thor and gracefully exit with an
error message.
Fixes https://github.com/bundler/bundler/issues/5358
(cherry picked from commit f93accce1e9d948eaef79e2f380fd7d396687808)
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Properly set native extension build args for git gems
Closes https://github.com/bundler/bundler/pull/5399
Closes https://github.com/bundler/bundler/issues/5401
(cherry picked from commit 74341aaaa949e914cd1cffb163cd99bdf14b91df)
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Permissive temporary home directory
We're deploying to a server with several Ruby applications each with it's own user. However, the home directory of these users is *not* writable by the users.
Since Bundler 1.14.x, a temporary home directoy in the `tmpdir` is used as a fallback. This works well for only one user, but will fail with more than one:
1. user1 execs `bundle`
2. `/tmp/bundler/home/user1` is created usually with 0755 modes
3. user2 execs `bundle`
4. `/tmp/bundler/home/user2` cannot be created since user2 can't write to `home`
This PR adds a `chmod` to apply mode 0777 to `home`. It rescues to nil in order to silently fail on step 4 (user2 cannot chmod `home` since it's owned by user1) as well as in case of any other problem performing chmod. Since we're in a fallback here, I guess it's okay to do this. However, if this is too much code smell, a `unless File.exist?` will do the trick as well.
(cherry picked from commit 311b090449b591d4fb96a725e548b4d35ac95ed0)
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Update vendored Molinillo to 0.5.6
See https://github.com/CocoaPods/Molinillo/releases/0.5.6
closes #5359
closes #5362
(cherry picked from commit 3360aac96610f640ddfef0d65303d15bc99539d9)
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
[UI::Silent] Ensure all the same methods as Shell are implemented
Finish up the work of #5368
(cherry picked from commit cad91be3c8f4ab5a71eee284c051cb2e7c2cb3a2)
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
[RubyVersion] Ensure passed-in versions are valid during init
Closes #5380
(cherry picked from commit 427f07f77f414f43d693c87c104546af3d9fe984)
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
r=segiddins
Fix silent.rb method interface
Hi,
This is my first contribution to bundler, so I started by covering bundler/friendly_errors.rb with unit tests.
After covering it, I spotted a bug in lib/bundler/ui/silent.rb, trace method interface, so I fix it in this PR.
Waiting for feedback!
(cherry picked from commit c0de8f625f0ae2ed5cc2b26673c875e2900a9829)
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Fix misleading warning message when a dependency is unused on any platform
The current warning message for a dep with empty platform is misleading/confusing. You see a sort of the below warning when you `bundle install`,
```
The dependency tzinfo-data (>= 0) will be unused by any of the platforms Bundler is installing for. Bundler is installing for ruby but the dependency is only for x86-mingw32, x86-mswin32, x64-mingw32, java. To add those platforms to the bundle, run `bundle lock --add-platform mingw, mswin, x64_mingw, jruby`.
```
However, there two issues exist in this message "run `bundle lock --add-platform mingw, mswin, x64_mingw, jruby`":
- You must not have comma (,) for separators
- these commas will be considered as comma itself. see [here](https://github.com/bundler/bundler/blob/242fcbcb2d15d39c63e3441756b77fbfcc1b73b2/lib/bundler/cli/lock.rb#L34-L41)
- mingw, mswin, x64_mingw, jruby are not valid platforms
- platforms must be specified as the formats that rubygems can understand, like (x86-mingw32, x86-mswin32, x64-mingw32, java)
This fixes these issues.
(cherry picked from commit cd872ef1e8281252786535df960275cd391933bc)
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
marketing-ify the post-install message
/cc @segiddins @mrb
(cherry picked from commit 5db7e161e9057603188600549c226a1b40807e6e)
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Show inject usage when args are passed incorrectly
Closes #5384
Now it shows the error :
```
$ bundle inject "gem_name" 1 "v"
ERROR: "bundle inject" was called with arguments ["gem_name", "1", "v"]
Usage: "bundle inject GEM VERSION"
$ bundle inject "gem_name" "v" 1
ERROR: "bundle inject" was called with arguments ["gem_name", "v", "1"]
Usage: "bundle inject GEM VERSION"
```
(cherry picked from commit ec5cab34d3b12467433ca1a1eafc460e42834084)
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Bundle show --outdated shouldn't be updating versions
This PR is for #5375. I have added a spec (currently failing) of the desired behavior for `bundle show --outdated`
I'm currently looking into a fix for this but welcome anyone else to come in as well.
(cherry picked from commit d9d699c174a06da8052bc7da0d5ba64438968653)
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
[FriendlyErrors] Update ISSUES.md URL
This URL no longer exists after the docs reorganization
\c @feministy
I'd like to release this as 1.14.x so people aren't frustrated when opening new issues
(cherry picked from commit fb8c5a7e509cb0a54a4b34f8a93b86484241a747)
|
| |\ \ \ \ \
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
r=indirect
[Definition] Fail when installing in frozen mode & gemspec deps have changed
Closes #5264
- [x] Specs
|
