| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Unfortunately, AWS S3 credentials may contain characters that are not
valid in a URI. That means that the credentials cannot be provided in
the URI "basic auth" format, which is the only format Bundler currently
supports. As a result, it seems that this feature is broken both in
Rubygems and here in Bundler. Attempts to fix the problem[1] have
stalled out, and no one on the Bundler core team uses S3 sources, so
we've decided to not release support for S3 sources as part of the
Bundler gem itself.
We're working on a plugin system that will allow S3 support to be
implemented and updated without requiring new versions of Bundler to be
released. In the meantime, if this support is very important to allow
you to continue to work, you can use the the 1.8.0.svn gem as a
workaround.
[1]: https://github.com/bundler/bundler/pull/3345
This reverts commit 1ded21eee6bd5d20a767f0b0e1a332d268788de2.
This reverts commit b98338344f0a2e33905aa7c4a573a289bf7de06f.
This reverts commit 26214a3fec5421c23888c8f65932af45b288a54b.
This reverts commit ec0ef95ac8444a4221ab0469ad9200e0eafa37e4.
|
|
|
|
|
|
|
|
|
|
|
|
| |
We're pulling out svn source support into a plugin. While we work on
that, you can use the 1.8.0.svn release, but subversion sources aren't
an official part of Bundler itself.
This reverts commit 049d281d0fdcc29297a21c6a66cd7efc38690675.
This reverts commit 22fecdd07fdf02edb1a8824fb73dd7e015507644.
This reverts commit 38f195e11f37ce5139af4ff3384eb2f26c2edb19.
This reverts commit 500436a33de0b884525dbf82cfc69332fc96f8b9.
This reverts commit 2c356be90a23921058cd14fd0e4a366da195021a.
|
| |
|
|
|
|
|
| |
old: -t [--test=set the test framework]
new: -t [--test] # set the test framework
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
As discussed in https://github.com/bundler/bundler/pull/3305#issuecomment-67024669, the gem generator should not generate default files that change the rights or responsibilities of gem authors without their explicit consent. Let's change the gem generator to ask gem authors what they want, and allow them to change it via config or --flags.
* Remove LICENSE from default gem
* Remove CODE_OF_CONDUCT from default gem
* Add --coc flag to generate CODE_OF_CONDUCT
* Add --mit flag to generate LICENSE
* On gem generation, if not set to true/false, ask user if they are willing to license their code permissively under the MIT license
* On gem generation, if not set to true/false, ask user if they are willing to add a Code of Conduct to their gem
* On gem generation, if not set to false/rspec/minitest, ask if the user would like to generate tests along with their gems. Save the answer as bundle config gem.tests.
* Ensure that --coc, --mit, --test flags overrule any config settings that may be set.
|
|
|
|
|
|
|
| |
this adds `package —cache-path`, supports `config cache_path foo`, and
honors the BUNDLE_CACHE_PATH environment variable.
closes #3351
|
|
|
|
|
|
|
|
|
| |
Rather than adding .consolerc and seeing the surface area of things that
Bundler supports continue to grow endlessly, Bundler 2.0 will deprecate
the `bundle console` command, and instead add a script to each generated
gem that loads `bundler/setup` and then loads the gem. Anyone can then
edit that file to run the REPL they want, load the fixtures they want,
and everything else, without anything specific to Bundler at all.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
we were previously forcing the configured auth if any was set, and it
was only by a quirk of the URI not matching if it had user and pass set
that we didn’t overwrite the Gemfile userinfo. Now that we also check
for hostname-level configured userinfo, we have to detect whether the
original URL in the Gemfile has userinfo and preserve it if it exists.
|
| |
|
|\
| |
| |
| |
| |
| |
| |
| |
| | |
Version 1.7.12
Conflicts:
CHANGELOG.md
lib/bundler/fetcher.rb
spec/cache/platform_spec.rb
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
It turns out that we already use Gemfile auth over configured auth,
because the Gemfile URI includes the username and password. That means
that mirrors and configured auth will not be used for any URI with auth
in it.
This reverts commit 88969b5210cbab4226cdcc4245762f8c3ba41f1a.
|
| | |
|
| | |
|
| |\
| | |
| | | |
Always use source credentials
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Previously, it tried to make anonymous requests, then retried with
credentials if it got a 401 or 403.
This caused issues with many private gem servers.
Fixes #3297, #3296, #3180.
|
| |/ |
|
| |
| |
| |
| |
| | |
This reverts commit 5d5651d35bbda4ad526c8f86bb2ce03f501a9eae, reversing
changes made to 714f23e99c9628260909e11099dcfba3dda35b0a.
|
| | |
|
| |\
| | |
| | | |
Ruby 2.2: support mri_22, (x64_)mingw_22 as platform options
|
| | |
| | |
| | |
| | |
| | |
| | | |
See e6bc48eb.
Signed-off-by: Alex Coles <alex@alexbcoles.com>
|
| |/
| |
| |
| |
| |
| | |
See e6bc48eb.
Signed-off-by: Alex Coles <alex@alexbcoles.com>
|
| | |
|
| | |
|
| |
| |
| |
| | |
This reverts commit 7f940849bd0e3ffb1a48433b5d058e51b0d3ae0f.
|
| |
| |
| |
| | |
This should fix the build on Ruby 1.8.7.
|
| | |
|
| | |
|
| |\
| | |
| | | |
Added ruby_22? and friends to CurrentRuby.
|
| |/ |
|
| |\
| | |
| | | |
Convert sources to lock format before comparing.
|
| |/
| |
| |
| |
| |
| |
| | |
If the Gemfile sources lock to the same result as the current locked
sources, they haven't changed.
Closes #3298.
|
| |
| |
| |
| | |
[ci skip]
|
|\ \
| | |
| | | |
untaint gem path loaded from github
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
require fails because security error, in condition of:
* $SAFE is 1
* gems from git or github. ex: gem 'bar', :github => 'foo/bar'
original full_gem_path method in rubygems is untaint all paths of gems (in find_full_gem_path method), but Bundler override the method without untaint.
|
|/ / |
|
| | |
|
| |
| |
| |
| | |
turns out Bundler.setup by itself won’t actually set the exitstatus
unless it’s also running inside a TTY (which this isn’t).
|
| | |
|
| |
| |
| |
| |
| |
| | |
apparently on some Ruby 1.8.7 installs, open3 doesn’t return an exit
status, and that includes all Travis installs of 1.8.7. :/ these tests
all pass (while checking exit status) on my machine, but they shouldn’t
fail if the Ruby on Travis isn’t able to provide exitstatuses.
|
| |
| |
| |
| |
| | |
I cannot figure out how to get it to pass with the new, faster exec,
but it passes in real life when I run the same command with (I think)
the same environment variables. Arrrrgh.
|