| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
| |
We can't use the same logic as "dependencies" for
"requirements". Because "," in "requirements" isn't escaped as "&".
|
|\
| |
| |
| |
| |
| |
| |
| | |
5521: bundler gem: Support test-unit as a testing framework r=hsbt a=kou
Co-authored-by: Kouhei Sutou <kou@clear-code.com>
|
| | |
|
|\ \
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
7529: Remove optional require net-http-pipeline from vendored net-http-persistent. r=deivid-rodriguez a=marciotoshio
### What was the end-user problem that led to this PR?
See issue https://github.com/bundler/bundler/issues/7521.
### What was your diagnosis of the problem?
My diagnosis was If you have net-http-pipeline gem installed bundler loads it and you can't vendor it.
### What is your fix for the problem, implemented in this PR?
My fix is remove the optional require of net-http-pipeline.
### Why did you choose this fix out of the possible options?
I chose this fix because was a recommendation in the issue by @deivid-rodriguez
Fixes #7521.
Co-authored-by: David Rodríguez <deivid.rodriguez@riseup.net>
Co-authored-by: Marcio Toshio <marcio@tosh.io>
|
| | | |
|
|/ /
| |
| |
| |
| | |
To avoid Gemfile resolution afterwards, and use the "inline
environment" instead.
|
| |
| |
| |
| |
| |
| |
| | |
So that it's always higher than the latest patch level release. This
saves us from having to merge back the version bump after each patch
level release, which is also not accurate since master is "bigger" than
the latest patch level release anyways.
|
| |
| |
| |
| |
| |
| | |
Previously, we were recommending `bundle config deployment true; bundle
install` as an alternative to `bundle install --deployment`, but they
were not working equally.
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This require causes circular require.
```
$ touch empty_file
$ RUBYGEMS_GEMDEPS=empty_file ./local/bin/ruby -w -e ''
/home/mame/work/ruby/local/lib/ruby/2.7.0/rubygems/core_ext/kernel_require.rb:92: warning: /home/mame/work/ruby/local/lib/ruby/2.7.0/rubygems/core_ext/kernel_require.rb:92: warning: loading in progress, circular require considered harmful - /home/mame/work/ruby/local/lib/ruby/2.7.0/rubygems.rb
from <internal:gem_prelude>:1:in `<internal:gem_prelude>'
from <internal:gem_prelude>:1:in `require'
from /home/mame/work/ruby/local/lib/ruby/2.7.0/rubygems.rb:1417:in `<top (required)>'
from /home/mame/work/ruby/local/lib/ruby/2.7.0/rubygems.rb:1203:in `use_gemdeps'
from /home/mame/work/ruby/local/lib/ruby/2.7.0/rubygems/user_interaction.rb:47:in `use_ui'
from /home/mame/work/ruby/local/lib/ruby/2.7.0/rubygems.rb:1204:in `block in use_gemdeps'
from /home/mame/work/ruby/local/lib/ruby/2.7.0/rubygems/core_ext/kernel_require.rb:92:in `require'
from /home/mame/work/ruby/local/lib/ruby/2.7.0/rubygems/core_ext/kernel_require.rb:92:in `require'
from /home/mame/work/ruby/local/lib/ruby/2.7.0/bundler.rb:11:in `<top (required)>'
from /home/mame/work/ruby/local/lib/ruby/2.7.0/bundler.rb:11:in `require_relative'
from /home/mame/work/ruby/local/lib/ruby/2.7.0/bundler/rubygems_integration.rb:3:in `<top (required)>'
from /home/mame/work/ruby/local/lib/ruby/2.7.0/rubygems/core_ext/kernel_require.rb:92:in `require'
from /home/mame/work/ruby/local/lib/ruby/2.7.0/rubygems/core_ext/kernel_require.rb:92:in `require'
```
|
|\ \
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
7510: Don't spawn a shell when git pushing on release r=deivid-rodriguez a=deivid-rodriguez
### What was the end-user problem that led to this PR?
The problem was that I almost went crazy when trying to release the last bundler version. `rake release` was hanging and I didn't know why.
### What was your diagnosis of the problem?
Thanks to passing `DEBUG=true` to the task, I noticed that it was `git push` and `git push --tags` commands that were hanging:
```
$ DEBUG=true bin/rake release
["gem", "build", "-V", "/home/deivid/Code/bundler/bundler.gemspec"]
bundler 2.1.2 built to pkg/bundler-2.1.2.gem.
["git", "diff", "--exit-code"]
["git", "diff-index", "--quiet", "--cached", "HEAD"]
["git", "tag"]
["git", "tag", "-m", "Version 2.1.2", "v2.1.2"]
Tagged v2.1.2.
git push git push --tags
^Crake aborted!
Interrupt:
/home/deivid/Code/bundler/lib/bundler/gem_helper.rb:198:in `read'
/home/deivid/Code/bundler/lib/bundler/gem_helper.rb:198:in `popen'
/home/deivid/Code/bundler/lib/bundler/gem_helper.rb:198:in `block in sh_with_status'
/home/deivid/Code/bundler/lib/bundler/shared_helpers.rb:52:in `chdir'
/home/deivid/Code/bundler/lib/bundler/shared_helpers.rb:52:in `block in chdir'
/home/deivid/Code/bundler/lib/bundler/shared_helpers.rb:51:in `chdir'
/home/deivid/Code/bundler/lib/bundler/gem_helper.rb:197:in `sh_with_status'
/home/deivid/Code/bundler/lib/bundler/gem_helper.rb:133:in `perform_git_push'
/home/deivid/Code/bundler/lib/bundler/gem_helper.rb:115:in `git_push'
/home/deivid/Code/bundler/lib/bundler/gem_helper.rb:64:in `block (2 levels) in install'
/home/deivid/Code/bundler/lib/bundler/gem_helper.rb:160:in `tag_version'
/home/deivid/Code/bundler/lib/bundler/gem_helper.rb:64:in `block in install'
/home/deivid/Code/bundler/Rakefile:14:in `block in invoke'
/home/deivid/Code/bundler/Rakefile:13:in `invoke'
/home/deivid/Code/bundler/spec/support/rubygems_ext.rb:87:in `load'
/home/deivid/Code/bundler/spec/support/rubygems_ext.rb:87:in `gem_load_and_activate'
/home/deivid/Code/bundler/spec/support/rubygems_ext.rb:45:in `gem_load'
Tasks: TOP => release => release:source_control_push
(See full trace by running task with --trace)
```
The debugging output is very interesting because it also tells us that these commands are the only ones passing Strings to `IO.popen` instead of Arrays. That means these commands are [spawning a new shell](https://ruby-doc.org/core-2.6.5/IO.html#method-c-popen).
That's when I realized that I have [hub](https://github.com/github/hub) installed on my system and `git` aliased to it. So I figure this aliasing was interacting with the subcommand in a bad way.
Indeed, disabling `hub` fixed the issue and let me make the release.
### What is your fix for the problem, implemented in this PR?
I think we can avoid other issues similar to this for people pushing releases by not spawing a shell here, just like we do with all of the other commands.
I think it's a good practice anyways.
### Why did you choose this fix out of the possible options?
I chose this fix because I think it makes the code better. Of course I can disable `hub` every time I release, but I think this is worth doing.
Co-authored-by: David Rodríguez <deivid.rodriguez@riseup.net>
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
These `git push` operations are the only ones passing a string argument
to `IO.popen` and those are the exact commands hanging for me when I
try to release. I'm not sure what the reason is (maybe that I use hub,
which does some fancy shell aliasing?), but I think it's a good practice
to avoid a new shell anyways.
|
| | | |
|
|\ \ \
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
7513: Improve error message when not logged in r=deivid-rodriguez a=steveklabnik
`gem push` makes you add other parameters, and is awkward to figure out the details. `gem signin` is a more direct way of doing this.
Co-authored-by: Steve Klabnik <steve@steveklabnik.com>
|
| |/ /
| | |
| | | |
`gem push` makes you add other parameters, and is awkward to figure out the details. `gem signin` is a more direct way of doing this.
|
|/ / |
|
| |
| |
| |
| |
| |
| | |
Some setups have `--disable=gems` in `RUBYOPT`, and sometimes use
`bundler` without modifying that env. Let's not break those setups just
to save a `require`.
|
| | |
|
|\ \
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
7498: Add compatibility methods for `rubygems-bundler` gem r=deivid-rodriguez a=deivid-rodriguez
### What was the end-user problem that led to this PR?
The problem was that `bundler install` on a `ruby` installation using `rvm` is broken by default.
### What was your diagnosis of the problem?
My diagnosis was that during 2.1.0 development phase I removed some stuff I thought it was internal and nobody would be using.
### What is your fix for the problem, implemented in this PR?
My fix is to restore a minimal compatibility layer so that this stuff works again.
### Why did you choose this fix out of the possible options?
I chose this fix because I think this is the minimum set of additions that gets stuff working again.
Fixes #7488.
Co-authored-by: David Rodríguez <deivid.rodriguez@riseup.net>
|
| | |
| | |
| | |
| | |
| | | |
Hopefully `rvm` won't install this gem by default, but for now
I'm adding the following two methods for compatibility with it.
|
|/ / |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We might be shelling out to rubygems in a `bundle exec` context. In the
case where we don't shell out to the `gem` binstub directly, the
previous trick wouldn't work.
Instead, reset the rubygems ui right after `bundler/setup`, so that if
we end up shelling out to rubygems, it will use its default shell
(non-silent).
The best place to fix this would probably be right inside the `gem`
script, but even if we fix it there, we'll need workarounds for previous
rubygems versions inside `bundler` so I think this is good for now.
|
| | |
|
| | |
|
|\ \
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
7460: Vendor `uri` library r=hsbt a=deivid-rodriguez
### What was the end-user problem that led to this PR?
The problem was that after the gemification of the `uri` library (which will happen in ruby 2.7), `bundler` will activate the default version of the new library inside its own `bundler/setup` code. That means users won't be able to specify the version of the library they want/need to use in their own Gemfiles.
### What was your diagnosis of the problem?
My diagnosis was that we should avoid using `uri` inside `bundler/setup` code.
### What is your fix for the problem, implemented in this PR?
My fix is to vendor the `uri` library, like we did with `fileutils`.
Co-authored-by: David Rodríguez <deivid.rodriguez@riseup.net>
|
| | | |
|
| | |
| | |
| | |
| | |
| | | |
We vendorize it as a dependency of `net-http-persistent`, so usages
inside `net-http-persistent` get automatically replaced by `automatiek`.
|
| | | |
|
| | | |
|
| | | |
|
| |/ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
These gem task checks for a specific string in the gem subcommand
output. However, if we are inside a bundled environment (`bundle exec
rake install` instead of `rake install`), the ruby process will require
`bundler/setup` first thing, and initialize a silent UI for rubygems. As
a result, the output string will be always empty and the condition for
success will always fail. So, even if installation succeeded, user will
always be notified of a failure like:
```
foo 1.0 built to pkg/foo-1.0.gem.
rake aborted!
Couldn't install gem, run `gem install /home/deivid/Code/bundler/tmp/1/bundled_app/pkg/foo-1.0.gem' for more detailed output
...
```
So, don't check for a specific string in the output and only rely on the
exit status.
|
| |
| |
| |
| | |
And use it consistently.
|
|/
|
|
| |
Like it is named in the other analogous methods.
|
| |
|
|
|
|
|
|
|
| |
I think the comment is misleading and it referred to rubygems, not
gemstash, because this code is about the client side of things. If this
is correct, we can drop it because we don't support any rubygems version
under 2.5.2, and the fix was introduced in rubygems 2.5.
|
| |
|
| |
|
| |
|
|\
| |
| |
| |
| |
| |
| |
| |
| |
| | |
7418: Restore previous BUNDLE_GEMFILE in bundler/inline r=deivid-rodriguez a=fatkodima
[Without spacing changes](https://github.com/bundler/bundler/pull/7418/files?w=1)
Closes #7159
Co-authored-by: fatkodima <fatkodima123@gmail.com>
|
| | |
|
| |
| |
| |
| |
| | |
Since cgi is now a default gem on ruby 2.7, we're getting some
unintended activations of the new default gem inside our specs.
|
|\ \
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
7455: Lazily load `open3` r=deivid-rodriguez a=deivid-rodriguez
### What was the end-user problem that led to this PR?
The problem was `open3` will be gemified in ruby 2.7, and since we use it inside `bundler`, we might activate a version causing a conflict with the user's choice.
### What was your diagnosis of the problem?
My diagnosis was that only loading it when needed should be better.
### What is your fix for the problem, implemented in this PR?
My fix is to lazily load it.
I expect this PR to fix [some of the errors](https://travis-ci.org/bundler/bundler/jobs/615940817) currently happening in our CI against ruby-head.
Co-authored-by: David Rodríguez <deivid.rodriguez@riseup.net>
|
| |/
| |
| |
| |
| | |
Since it's a default gem now, it's better to load it as late as
possible.
|
|/ |
|
|\
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
7442: Fix `bundle exec`'ing to rubygems being silent r=deivid-rodriguez a=deivid-rodriguez
### What was the end-user problem that led to this PR?
The problem was that #7401 caused a regression where `bundle exec`'ing to rubygems would silence rubygems output.
### What was your diagnosis of the problem?
My diagnosis was that the removal of:
* Code where `Bundler::RGProxy` would be only set conditionally on `Bundler.ui =` not being passed `nil`.
* Code setting `Bundler.ui` to `nil` right before shelling during `bundle exec`.
caused rubygems UI to be silent during `bundle exec gem`.
### What is your fix for the problem, implemented in this PR?
My fix is to explictly "unsilence" rubygems UI before `bundle exec` calls.
### Why did you choose this fix out of the possible options?
I chose this fix because it's more explicit than the previous one.
Fixes #7441.
Co-authored-by: David Rodríguez <deivid.rodriguez@riseup.net>
|
| | |
|
|/ |
|
|
|
|
|
|
| |
Using `bin/rake vendor:thor[master]`.
So that we're ready for an upcoming release.
|