2018-11-02 19:07:56 -04:00
# frozen_string_literal: true
RSpec . describe " bundle exec " do
let ( :system_gems_to_install ) { %w[ rack-1.0.0 rack-0.9.1 ] }
before :each do
system_gems ( system_gems_to_install , :path = > :bundle_path )
end
it " works with --gemfile flag " do
create_file " CustomGemfile " , <<-G
gem " rack " , " 1.0.0 "
G
bundle " exec --gemfile CustomGemfile rackup "
expect ( out ) . to eq ( " 1.0.0 " )
end
it " activates the correct gem " do
gemfile <<-G
gem " rack " , " 0.9.1 "
G
bundle " exec rackup "
expect ( out ) . to eq ( " 0.9.1 " )
end
it " works when the bins are in ~/.bundle " do
install_gemfile <<-G
gem " rack "
G
bundle " exec rackup "
expect ( out ) . to eq ( " 1.0.0 " )
end
Fix some bundler specs (#2380)
* These seem to consistenly pass already
* Show actual command when running `make test-bundler`
Current the setup command that installs the necessary gems for testing
bundler was printed, but not the actual command that runs the tests.
That was a bit confusing.
* Borrow trick from setproctitle specs
* A title that long doesn't get set sometimes
No idea why, but the test doesn't need that the title is that long.
* Fix most gem helper spec ruby-core failures
* Fix the rest of the gem helper failures
* Fix version spec by improving the assertion
* Remove unnecessary `BUNDLE_RUBY` environment var
We can use `RUBY` when necessary, and `BUNDLE_RUBY` is not a good name
because bundler considers `BUNDLE_*` variables as settings.
* Rename `BUNDLE_GEM` to `GEM_COMMAND`
This is more descriptive I think, and also friendlier for bundler
because `BUNDLE_` env variables are interpreted by bundler as settings,
and this is not a bundler setting.
This fixes one bundler spec failure in config specs against ruby-core.
* Fix quality spec when run in core
Use the proper path helper.
* Fix dummy lib builder to never load default gems
If a dummy library is named as a default gem, when requiring the library
from its executable, the default gem would be loaded when running from
core, because in core all default gems share path with bundler, and thus
they are always in the $LOAD_PATH. We fix the issue by loading lib
relatively inside dummy lib executables.
* More exact assertions
Sometimes I have the problem that I do some "print debugging" inside
specs, and suddently the spec passes. This happens when the assertion is
too relaxed, and the things I print make it match, specially when they
are simple strings like "1.0" than can be easily be part of gem paths
that I print for debugging.
I fix this by making a more exact assertion.
* Detect the correct shebang when ENV["RUBY"] is set
* Relax assertion
So that the spec passes even if another paths containing "ext" are in
the load path. This works to fix a ruby-core issue, but it's a better
assertion in general. We just want to know that the extension path was
added.
* Use folder structure independent path helper
It should fix this spec for ruby-core.
* Fix the last failing spec on ruby-core
* Skip `bundle open <default_gem>` spec when no default gems
2019-08-19 20:46:31 -04:00
it " works when running from a random directory " do
2018-11-02 19:07:56 -04:00
install_gemfile <<-G
gem " rack "
G
2019-04-14 02:01:35 -04:00
bundle " exec 'cd #{ tmp ( " gems " ) } && rackup' "
2018-11-02 19:07:56 -04:00
2019-04-26 13:02:19 -04:00
expect ( out ) . to eq ( " 1.0.0 " )
2018-11-02 19:07:56 -04:00
end
it " works when exec'ing something else " do
install_gemfile 'gem "rack"'
bundle " exec echo exec "
expect ( out ) . to eq ( " exec " )
end
it " works when exec'ing to ruby " do
install_gemfile 'gem "rack"'
2019-04-14 02:01:35 -04:00
bundle " exec ruby -e 'puts %{hi}' "
2018-11-02 19:07:56 -04:00
expect ( out ) . to eq ( " hi " )
end
2019-08-23 04:18:37 -04:00
it " respects custom process title when loading through ruby " do
Fix some bundler specs (#2380)
* These seem to consistenly pass already
* Show actual command when running `make test-bundler`
Current the setup command that installs the necessary gems for testing
bundler was printed, but not the actual command that runs the tests.
That was a bit confusing.
* Borrow trick from setproctitle specs
* A title that long doesn't get set sometimes
No idea why, but the test doesn't need that the title is that long.
* Fix most gem helper spec ruby-core failures
* Fix the rest of the gem helper failures
* Fix version spec by improving the assertion
* Remove unnecessary `BUNDLE_RUBY` environment var
We can use `RUBY` when necessary, and `BUNDLE_RUBY` is not a good name
because bundler considers `BUNDLE_*` variables as settings.
* Rename `BUNDLE_GEM` to `GEM_COMMAND`
This is more descriptive I think, and also friendlier for bundler
because `BUNDLE_` env variables are interpreted by bundler as settings,
and this is not a bundler setting.
This fixes one bundler spec failure in config specs against ruby-core.
* Fix quality spec when run in core
Use the proper path helper.
* Fix dummy lib builder to never load default gems
If a dummy library is named as a default gem, when requiring the library
from its executable, the default gem would be loaded when running from
core, because in core all default gems share path with bundler, and thus
they are always in the $LOAD_PATH. We fix the issue by loading lib
relatively inside dummy lib executables.
* More exact assertions
Sometimes I have the problem that I do some "print debugging" inside
specs, and suddently the spec passes. This happens when the assertion is
too relaxed, and the things I print make it match, specially when they
are simple strings like "1.0" than can be easily be part of gem paths
that I print for debugging.
I fix this by making a more exact assertion.
* Detect the correct shebang when ENV["RUBY"] is set
* Relax assertion
So that the spec passes even if another paths containing "ext" are in
the load path. This works to fix a ruby-core issue, but it's a better
assertion in general. We just want to know that the extension path was
added.
* Use folder structure independent path helper
It should fix this spec for ruby-core.
* Fix the last failing spec on ruby-core
* Skip `bundle open <default_gem>` spec when no default gems
2019-08-19 20:46:31 -04:00
script_that_changes_its_own_title_and_checks_if_picked_up_by_ps_unix_utility = << ~ 'RUBY'
Process . setproctitle ( " 1-2-3-4-5-6-7-8-9-10-11-12-13-14-15 " )
puts ` ps -ocommand= -p #{ $$ } `
2019-06-01 05:49:40 -04:00
RUBY
create_file " Gemfile "
create_file " a.rb " , script_that_changes_its_own_title_and_checks_if_picked_up_by_ps_unix_utility
bundle " exec ruby a.rb "
Fix some bundler specs (#2380)
* These seem to consistenly pass already
* Show actual command when running `make test-bundler`
Current the setup command that installs the necessary gems for testing
bundler was printed, but not the actual command that runs the tests.
That was a bit confusing.
* Borrow trick from setproctitle specs
* A title that long doesn't get set sometimes
No idea why, but the test doesn't need that the title is that long.
* Fix most gem helper spec ruby-core failures
* Fix the rest of the gem helper failures
* Fix version spec by improving the assertion
* Remove unnecessary `BUNDLE_RUBY` environment var
We can use `RUBY` when necessary, and `BUNDLE_RUBY` is not a good name
because bundler considers `BUNDLE_*` variables as settings.
* Rename `BUNDLE_GEM` to `GEM_COMMAND`
This is more descriptive I think, and also friendlier for bundler
because `BUNDLE_` env variables are interpreted by bundler as settings,
and this is not a bundler setting.
This fixes one bundler spec failure in config specs against ruby-core.
* Fix quality spec when run in core
Use the proper path helper.
* Fix dummy lib builder to never load default gems
If a dummy library is named as a default gem, when requiring the library
from its executable, the default gem would be loaded when running from
core, because in core all default gems share path with bundler, and thus
they are always in the $LOAD_PATH. We fix the issue by loading lib
relatively inside dummy lib executables.
* More exact assertions
Sometimes I have the problem that I do some "print debugging" inside
specs, and suddently the spec passes. This happens when the assertion is
too relaxed, and the things I print make it match, specially when they
are simple strings like "1.0" than can be easily be part of gem paths
that I print for debugging.
I fix this by making a more exact assertion.
* Detect the correct shebang when ENV["RUBY"] is set
* Relax assertion
So that the spec passes even if another paths containing "ext" are in
the load path. This works to fix a ruby-core issue, but it's a better
assertion in general. We just want to know that the extension path was
added.
* Use folder structure independent path helper
It should fix this spec for ruby-core.
* Fix the last failing spec on ruby-core
* Skip `bundle open <default_gem>` spec when no default gems
2019-08-19 20:46:31 -04:00
expect ( out ) . to eq ( " 1-2-3-4-5-6-7-8-9-10-11-12-13-14-15 " )
2019-06-01 05:49:40 -04:00
end
2018-11-02 19:07:56 -04:00
it " accepts --verbose " do
install_gemfile 'gem "rack"'
bundle " exec --verbose echo foobar "
expect ( out ) . to eq ( " foobar " )
end
it " passes --verbose to command if it is given after the command " do
install_gemfile 'gem "rack"'
bundle " exec echo --verbose "
expect ( out ) . to eq ( " --verbose " )
end
it " handles --keep-file-descriptors " do
require " tempfile "
command = Tempfile . new ( " io-test " )
command . sync = true
command . write <<-G
if ARGV [ 0 ]
IO . for_fd ( ARGV [ 0 ] . to_i )
else
require 'tempfile'
io = Tempfile . new ( " io-test-fd " )
2019-11-11 03:57:45 -05:00
args = %W[ #{ Gem . ruby } -I #{ lib_dir } #{ bindir . join ( " bundle " ) } exec --keep-file-descriptors #{ Gem . ruby } #{ command . path } \# {io.to_i} ]
2019-04-14 02:01:35 -04:00
args << { io . to_i = > io }
2018-11-02 19:07:56 -04:00
exec ( * args )
end
G
install_gemfile " "
2019-04-14 02:01:35 -04:00
sys_exec " #{ Gem . ruby } #{ command . path } "
2018-11-02 19:07:56 -04:00
2019-04-14 02:01:35 -04:00
expect ( out ) . to eq ( " " )
2019-06-01 05:49:40 -04:00
expect ( err ) . to be_empty
2018-11-02 19:07:56 -04:00
end
it " accepts --keep-file-descriptors " do
install_gemfile " "
bundle " exec --keep-file-descriptors echo foobar "
2019-06-01 05:49:40 -04:00
expect ( err ) . to be_empty
2018-11-02 19:07:56 -04:00
end
it " can run a command named --verbose " do
install_gemfile 'gem "rack"'
File . open ( " --verbose " , " w " ) do | f |
f . puts " # !/bin/sh "
f . puts " echo foobar "
end
File . chmod ( 0 o744 , " --verbose " )
with_path_as ( " . " ) do
bundle " exec -- --verbose "
end
expect ( out ) . to eq ( " foobar " )
end
it " handles different versions in different bundles " do
build_repo2 do
build_gem " rack_two " , " 1.0.0 " do | s |
s . executables = " rackup "
end
end
install_gemfile <<-G
2019-05-06 12:06:21 -04:00
source " #{ file_uri_for ( gem_repo1 ) } "
2018-11-02 19:07:56 -04:00
gem " rack " , " 0.9.1 "
G
Dir . chdir bundled_app2 do
install_gemfile bundled_app2 ( " Gemfile " ) , <<-G
2019-05-06 12:06:21 -04:00
source " #{ file_uri_for ( gem_repo2 ) } "
2018-11-02 19:07:56 -04:00
gem " rack_two " , " 1.0.0 "
G
end
bundle! " exec rackup "
expect ( out ) . to eq ( " 0.9.1 " )
Dir . chdir bundled_app2 do
bundle! " exec rackup "
expect ( out ) . to eq ( " 1.0.0 " )
end
end
2019-04-14 02:01:35 -04:00
context " with default gems " do
let ( :system_gems_to_install ) { [ ] }
let ( :default_irb_version ) { ruby " gem 'irb', '< 999999'; require 'irb'; puts IRB::VERSION " }
context " when not specified in Gemfile " do
before do
skip " irb isn't a default gem " if default_irb_version . empty?
install_gemfile " "
end
it " uses version provided by ruby " do
bundle! " exec irb --version "
expect ( out ) . to include ( default_irb_version )
2019-06-01 05:49:40 -04:00
expect ( err ) . to be_empty
2019-04-14 02:01:35 -04:00
end
end
context " when specified in Gemfile directly " do
let ( :specified_irb_version ) { " 0.9.6 " }
before do
skip " irb isn't a default gem " if default_irb_version . empty?
build_repo2 do
build_gem " irb " , specified_irb_version do | s |
s . executables = " irb "
end
end
install_gemfile <<-G
2019-05-06 12:06:21 -04:00
source " #{ file_uri_for ( gem_repo2 ) } "
2019-04-14 02:01:35 -04:00
gem " irb " , " #{ specified_irb_version } "
G
end
it " uses version specified " do
bundle! " exec irb --version "
Fix some bundler specs (#2380)
* These seem to consistenly pass already
* Show actual command when running `make test-bundler`
Current the setup command that installs the necessary gems for testing
bundler was printed, but not the actual command that runs the tests.
That was a bit confusing.
* Borrow trick from setproctitle specs
* A title that long doesn't get set sometimes
No idea why, but the test doesn't need that the title is that long.
* Fix most gem helper spec ruby-core failures
* Fix the rest of the gem helper failures
* Fix version spec by improving the assertion
* Remove unnecessary `BUNDLE_RUBY` environment var
We can use `RUBY` when necessary, and `BUNDLE_RUBY` is not a good name
because bundler considers `BUNDLE_*` variables as settings.
* Rename `BUNDLE_GEM` to `GEM_COMMAND`
This is more descriptive I think, and also friendlier for bundler
because `BUNDLE_` env variables are interpreted by bundler as settings,
and this is not a bundler setting.
This fixes one bundler spec failure in config specs against ruby-core.
* Fix quality spec when run in core
Use the proper path helper.
* Fix dummy lib builder to never load default gems
If a dummy library is named as a default gem, when requiring the library
from its executable, the default gem would be loaded when running from
core, because in core all default gems share path with bundler, and thus
they are always in the $LOAD_PATH. We fix the issue by loading lib
relatively inside dummy lib executables.
* More exact assertions
Sometimes I have the problem that I do some "print debugging" inside
specs, and suddently the spec passes. This happens when the assertion is
too relaxed, and the things I print make it match, specially when they
are simple strings like "1.0" than can be easily be part of gem paths
that I print for debugging.
I fix this by making a more exact assertion.
* Detect the correct shebang when ENV["RUBY"] is set
* Relax assertion
So that the spec passes even if another paths containing "ext" are in
the load path. This works to fix a ruby-core issue, but it's a better
assertion in general. We just want to know that the extension path was
added.
* Use folder structure independent path helper
It should fix this spec for ruby-core.
* Fix the last failing spec on ruby-core
* Skip `bundle open <default_gem>` spec when no default gems
2019-08-19 20:46:31 -04:00
expect ( out ) . to eq ( specified_irb_version )
2019-06-01 05:49:40 -04:00
expect ( err ) . to be_empty
2019-04-14 02:01:35 -04:00
end
end
context " when specified in Gemfile indirectly " do
let ( :indirect_irb_version ) { " 0.9.6 " }
before do
skip " irb isn't a default gem " if default_irb_version . empty?
build_repo2 do
build_gem " irb " , indirect_irb_version do | s |
s . executables = " irb "
end
build_gem " gem_depending_on_old_irb " do | s |
s . add_dependency " irb " , indirect_irb_version
end
end
install_gemfile <<-G
2019-05-06 12:06:21 -04:00
source " #{ file_uri_for ( gem_repo2 ) } "
2019-04-14 02:01:35 -04:00
gem " gem_depending_on_old_irb "
G
bundle! " exec irb --version "
end
it " uses resolved version " do
Fix some bundler specs (#2380)
* These seem to consistenly pass already
* Show actual command when running `make test-bundler`
Current the setup command that installs the necessary gems for testing
bundler was printed, but not the actual command that runs the tests.
That was a bit confusing.
* Borrow trick from setproctitle specs
* A title that long doesn't get set sometimes
No idea why, but the test doesn't need that the title is that long.
* Fix most gem helper spec ruby-core failures
* Fix the rest of the gem helper failures
* Fix version spec by improving the assertion
* Remove unnecessary `BUNDLE_RUBY` environment var
We can use `RUBY` when necessary, and `BUNDLE_RUBY` is not a good name
because bundler considers `BUNDLE_*` variables as settings.
* Rename `BUNDLE_GEM` to `GEM_COMMAND`
This is more descriptive I think, and also friendlier for bundler
because `BUNDLE_` env variables are interpreted by bundler as settings,
and this is not a bundler setting.
This fixes one bundler spec failure in config specs against ruby-core.
* Fix quality spec when run in core
Use the proper path helper.
* Fix dummy lib builder to never load default gems
If a dummy library is named as a default gem, when requiring the library
from its executable, the default gem would be loaded when running from
core, because in core all default gems share path with bundler, and thus
they are always in the $LOAD_PATH. We fix the issue by loading lib
relatively inside dummy lib executables.
* More exact assertions
Sometimes I have the problem that I do some "print debugging" inside
specs, and suddently the spec passes. This happens when the assertion is
too relaxed, and the things I print make it match, specially when they
are simple strings like "1.0" than can be easily be part of gem paths
that I print for debugging.
I fix this by making a more exact assertion.
* Detect the correct shebang when ENV["RUBY"] is set
* Relax assertion
So that the spec passes even if another paths containing "ext" are in
the load path. This works to fix a ruby-core issue, but it's a better
assertion in general. We just want to know that the extension path was
added.
* Use folder structure independent path helper
It should fix this spec for ruby-core.
* Fix the last failing spec on ruby-core
* Skip `bundle open <default_gem>` spec when no default gems
2019-08-19 20:46:31 -04:00
expect ( out ) . to eq ( indirect_irb_version )
2019-06-01 05:49:40 -04:00
expect ( err ) . to be_empty
2019-04-14 02:01:35 -04:00
end
end
end
it " warns about executable conflicts " do
build_repo2 do
build_gem " rack_two " , " 1.0.0 " do | s |
s . executables = " rackup "
end
end
bundle " config set path.system true "
install_gemfile <<-G
2019-05-06 12:06:21 -04:00
source " #{ file_uri_for ( gem_repo1 ) } "
2019-04-14 02:01:35 -04:00
gem " rack " , " 0.9.1 "
G
Dir . chdir bundled_app2 do
install_gemfile bundled_app2 ( " Gemfile " ) , <<-G
2019-05-06 12:06:21 -04:00
source " #{ file_uri_for ( gem_repo2 ) } "
2019-04-14 02:01:35 -04:00
gem " rack_two " , " 1.0.0 "
G
end
bundle! " exec rackup "
expect ( last_command . stderr ) . to eq (
" Bundler is using a binstub that was created for a different gem (rack). \n " \
" You should run `bundle binstub rack_two` to work around a system/bundle conflict. "
)
end
2018-11-02 19:07:56 -04:00
it " handles gems installed with --without " do
install_gemfile <<-G, forgotten_command_line_options(:without => "middleware")
2019-05-06 12:06:21 -04:00
source " #{ file_uri_for ( gem_repo1 ) } "
2018-11-02 19:07:56 -04:00
gem " rack " # rack 0.9.1 and 1.0 exist
group :middleware do
gem " rack_middleware " # rack_middleware depends on rack 0.9.1
end
G
bundle " exec rackup "
expect ( out ) . to eq ( " 0.9.1 " )
expect ( the_bundle ) . not_to include_gems " rack_middleware 1.0 "
end
it " does not duplicate already exec'ed RUBYOPT " do
install_gemfile <<-G
gem " rack "
G
rubyopt = ENV [ " RUBYOPT " ]
2019-11-11 03:57:45 -05:00
rubyopt = " -r #{ lib_dir } /bundler/setup #{ rubyopt } "
2018-11-02 19:07:56 -04:00
bundle " exec 'echo $RUBYOPT' "
expect ( out ) . to have_rubyopts ( rubyopt )
bundle " exec 'echo $RUBYOPT' " , :env = > { " RUBYOPT " = > rubyopt }
expect ( out ) . to have_rubyopts ( rubyopt )
end
it " does not duplicate already exec'ed RUBYLIB " do
install_gemfile <<-G
gem " rack "
G
rubylib = ENV [ " RUBYLIB " ]
2019-11-11 03:57:45 -05:00
rubylib = rubylib . to_s . split ( File :: PATH_SEPARATOR ) . unshift lib_dir . to_s
2018-11-02 19:07:56 -04:00
rubylib = rubylib . uniq . join ( File :: PATH_SEPARATOR )
bundle " exec 'echo $RUBYLIB' "
expect ( out ) . to include ( rubylib )
bundle " exec 'echo $RUBYLIB' " , :env = > { " RUBYLIB " = > rubylib }
expect ( out ) . to include ( rubylib )
end
it " errors nicely when the argument doesn't exist " do
install_gemfile <<-G
gem " rack "
G
bundle " exec foobarbaz "
expect ( exitstatus ) . to eq ( 127 ) if exitstatus
2019-04-14 02:01:35 -04:00
expect ( err ) . to include ( " bundler: command not found: foobarbaz " )
expect ( err ) . to include ( " Install missing gem executables with `bundle install` " )
2018-11-02 19:07:56 -04:00
end
it " errors nicely when the argument is not executable " do
install_gemfile <<-G
gem " rack "
G
bundle " exec touch foo "
bundle " exec ./foo "
expect ( exitstatus ) . to eq ( 126 ) if exitstatus
2019-04-14 02:01:35 -04:00
expect ( err ) . to include ( " bundler: not executable: ./foo " )
2018-11-02 19:07:56 -04:00
end
it " errors nicely when no arguments are passed " do
install_gemfile <<-G
gem " rack "
G
bundle " exec "
expect ( exitstatus ) . to eq ( 128 ) if exitstatus
2019-04-14 02:01:35 -04:00
expect ( err ) . to include ( " bundler: exec needs a command to run " )
2018-11-02 19:07:56 -04:00
end
Fix some bundler specs (#2380)
* These seem to consistenly pass already
* Show actual command when running `make test-bundler`
Current the setup command that installs the necessary gems for testing
bundler was printed, but not the actual command that runs the tests.
That was a bit confusing.
* Borrow trick from setproctitle specs
* A title that long doesn't get set sometimes
No idea why, but the test doesn't need that the title is that long.
* Fix most gem helper spec ruby-core failures
* Fix the rest of the gem helper failures
* Fix version spec by improving the assertion
* Remove unnecessary `BUNDLE_RUBY` environment var
We can use `RUBY` when necessary, and `BUNDLE_RUBY` is not a good name
because bundler considers `BUNDLE_*` variables as settings.
* Rename `BUNDLE_GEM` to `GEM_COMMAND`
This is more descriptive I think, and also friendlier for bundler
because `BUNDLE_` env variables are interpreted by bundler as settings,
and this is not a bundler setting.
This fixes one bundler spec failure in config specs against ruby-core.
* Fix quality spec when run in core
Use the proper path helper.
* Fix dummy lib builder to never load default gems
If a dummy library is named as a default gem, when requiring the library
from its executable, the default gem would be loaded when running from
core, because in core all default gems share path with bundler, and thus
they are always in the $LOAD_PATH. We fix the issue by loading lib
relatively inside dummy lib executables.
* More exact assertions
Sometimes I have the problem that I do some "print debugging" inside
specs, and suddently the spec passes. This happens when the assertion is
too relaxed, and the things I print make it match, specially when they
are simple strings like "1.0" than can be easily be part of gem paths
that I print for debugging.
I fix this by making a more exact assertion.
* Detect the correct shebang when ENV["RUBY"] is set
* Relax assertion
So that the spec passes even if another paths containing "ext" are in
the load path. This works to fix a ruby-core issue, but it's a better
assertion in general. We just want to know that the extension path was
added.
* Use folder structure independent path helper
It should fix this spec for ruby-core.
* Fix the last failing spec on ruby-core
* Skip `bundle open <default_gem>` spec when no default gems
2019-08-19 20:46:31 -04:00
it " raises a helpful error when exec'ing to something outside of the bundle " do
2019-04-14 02:01:35 -04:00
bundle! " config set clean false " # want to keep the rackup binstub
2018-11-02 19:07:56 -04:00
install_gemfile! <<-G
2019-05-06 12:06:21 -04:00
source " #{ file_uri_for ( gem_repo1 ) } "
2018-11-02 19:07:56 -04:00
gem " with_license "
G
[ true , false ] . each do | l |
2019-04-14 02:01:35 -04:00
bundle! " config set disable_exec_load #{ l } "
2018-11-02 19:07:56 -04:00
bundle " exec rackup "
2019-06-01 05:49:40 -04:00
expect ( err ) . to include " can't find executable rackup for gem rack. rack is not currently included in the bundle, perhaps you meant to add it to your Gemfile? "
2018-11-02 19:07:56 -04:00
end
end
describe " with help flags " do
each_prefix = proc do | string , & blk |
1 . upto ( string . length ) { | l | blk . call ( string [ 0 , l ] ) }
end
each_prefix . call ( " exec " ) do | exec |
describe " when #{ exec } is used " do
before ( :each ) do
install_gemfile <<-G
gem " rack "
G
create_file ( " print_args " , <<-'RUBY')
#!/usr/bin/env ruby
puts " args: #{ ARGV . inspect } "
RUBY
bundled_app ( " print_args " ) . chmod ( 0 o755 )
end
it " shows executable's man page when --help is after the executable " do
bundle " #{ exec } print_args --help "
expect ( out ) . to eq ( 'args: ["--help"]' )
end
it " shows executable's man page when --help is after the executable and an argument " do
bundle " #{ exec } print_args foo --help "
expect ( out ) . to eq ( 'args: ["foo", "--help"]' )
bundle " #{ exec } print_args foo bar --help "
expect ( out ) . to eq ( 'args: ["foo", "bar", "--help"]' )
bundle " #{ exec } print_args foo --help bar "
expect ( out ) . to eq ( 'args: ["foo", "--help", "bar"]' )
end
it " shows executable's man page when the executable has a - " do
FileUtils . mv ( bundled_app ( " print_args " ) , bundled_app ( " docker-template " ) )
bundle " #{ exec } docker-template build discourse --help "
expect ( out ) . to eq ( 'args: ["build", "discourse", "--help"]' )
end
it " shows executable's man page when --help is after another flag " do
bundle " #{ exec } print_args --bar --help "
expect ( out ) . to eq ( 'args: ["--bar", "--help"]' )
end
it " uses executable's original behavior for -h " do
bundle " #{ exec } print_args -h "
expect ( out ) . to eq ( 'args: ["-h"]' )
end
it " shows bundle-exec's man page when --help is between exec and the executable " do
with_fake_man do
bundle " #{ exec } --help cat "
end
expect ( out ) . to include ( %( [" #{ root } /man/bundle-exec.1"] ) )
end
it " shows bundle-exec's man page when --help is before exec " do
with_fake_man do
bundle " --help #{ exec } "
end
expect ( out ) . to include ( %( [" #{ root } /man/bundle-exec.1"] ) )
end
it " shows bundle-exec's man page when -h is before exec " do
with_fake_man do
bundle " -h #{ exec } "
end
expect ( out ) . to include ( %( [" #{ root } /man/bundle-exec.1"] ) )
end
it " shows bundle-exec's man page when --help is after exec " do
with_fake_man do
bundle " #{ exec } --help "
end
expect ( out ) . to include ( %( [" #{ root } /man/bundle-exec.1"] ) )
end
it " shows bundle-exec's man page when -h is after exec " do
with_fake_man do
bundle " #{ exec } -h "
end
expect ( out ) . to include ( %( [" #{ root } /man/bundle-exec.1"] ) )
end
end
end
end
describe " with gem executables " do
describe " run from a random directory " , :ruby_repo do
before ( :each ) do
install_gemfile <<-G
gem " rack "
G
end
it " works when unlocked " do
2019-04-14 02:01:35 -04:00
bundle " exec 'cd #{ tmp ( " gems " ) } && rackup' "
2018-11-02 19:07:56 -04:00
expect ( out ) . to eq ( " 1.0.0 " )
end
it " works when locked " do
expect ( the_bundle ) . to be_locked
2019-04-14 02:01:35 -04:00
bundle " exec 'cd #{ tmp ( " gems " ) } && rackup' "
2019-04-26 13:02:19 -04:00
expect ( out ) . to eq ( " 1.0.0 " )
2018-11-02 19:07:56 -04:00
end
end
describe " from gems bundled via :path " do
before ( :each ) do
build_lib " fizz " , :path = > home ( " fizz " ) do | s |
s . executables = " fizz "
end
install_gemfile <<-G
gem " fizz " , :path = > " #{ File . expand_path ( home ( " fizz " ) ) } "
G
end
it " works when unlocked " do
bundle " exec fizz "
expect ( out ) . to eq ( " 1.0 " )
end
it " works when locked " do
expect ( the_bundle ) . to be_locked
bundle " exec fizz "
expect ( out ) . to eq ( " 1.0 " )
end
end
describe " from gems bundled via :git " do
before ( :each ) do
build_git " fizz_git " do | s |
s . executables = " fizz_git "
end
install_gemfile <<-G
gem " fizz_git " , :git = > " #{ lib_path ( " fizz_git-1.0 " ) } "
G
end
it " works when unlocked " do
bundle " exec fizz_git "
expect ( out ) . to eq ( " 1.0 " )
end
it " works when locked " do
expect ( the_bundle ) . to be_locked
bundle " exec fizz_git "
expect ( out ) . to eq ( " 1.0 " )
end
end
describe " from gems bundled via :git with no gemspec " do
before ( :each ) do
build_git " fizz_no_gemspec " , :gemspec = > false do | s |
s . executables = " fizz_no_gemspec "
end
install_gemfile <<-G
gem " fizz_no_gemspec " , " 1.0 " , :git = > " #{ lib_path ( " fizz_no_gemspec-1.0 " ) } "
G
end
it " works when unlocked " do
bundle " exec fizz_no_gemspec "
expect ( out ) . to eq ( " 1.0 " )
end
it " works when locked " do
expect ( the_bundle ) . to be_locked
bundle " exec fizz_no_gemspec "
expect ( out ) . to eq ( " 1.0 " )
end
end
end
it " performs an automatic bundle install " do
gemfile <<-G
2019-05-06 12:06:21 -04:00
source " #{ file_uri_for ( gem_repo1 ) } "
2018-11-02 19:07:56 -04:00
gem " rack " , " 0.9.1 "
gem " foo "
G
2019-04-14 02:01:35 -04:00
bundle " config set auto_install 1 "
2018-11-02 19:07:56 -04:00
bundle " exec rackup "
expect ( out ) . to include ( " Installing foo 1.0 " )
end
2019-04-14 02:05:27 -04:00
describe " with gems bundled via :path with invalid gemspecs " do
2019-04-14 02:01:35 -04:00
it " outputs the gemspec validation errors " do
2018-11-02 19:07:56 -04:00
build_lib " foo "
gemspec = lib_path ( " foo-1.0 " ) . join ( " foo.gemspec " ) . to_s
File . open ( gemspec , " w " ) do | f |
f . write <<-G
Gem :: Specification . new do | s |
s . name = 'foo'
s . version = '1.0'
s . summary = 'TODO: Add summary'
s . authors = 'Me'
end
G
end
install_gemfile <<-G
gem " foo " , :path = > " #{ lib_path ( " foo-1.0 " ) } "
G
bundle " exec irb "
2019-06-01 05:49:40 -04:00
expect ( err ) . to match ( " The gemspec at #{ lib_path ( " foo-1.0 " ) . join ( " foo.gemspec " ) } is not valid " )
expect ( err ) . to match ( '"TODO" is not a summary' )
2018-11-02 19:07:56 -04:00
end
end
describe " with gems bundled for deployment " do
it " works when calling bundler from another script " do
gemfile <<-G
module Monkey
def bin_path ( a , b , c )
raise Gem :: GemNotFoundException . new ( 'Fail' )
end
end
Bundler . rubygems . extend ( Monkey )
G
bundle " install --deployment "
2019-04-14 02:01:35 -04:00
bundle " exec ruby -e '` #{ bindir . join ( " bundler " ) } -v`; puts $?.success?' "
2018-11-02 19:07:56 -04:00
expect ( out ) . to match ( " true " )
end
end
context " `load`ing a ruby file instead of `exec`ing " do
let ( :path ) { bundled_app ( " ruby_executable " ) }
let ( :shebang ) { " # !/usr/bin/env ruby " }
let ( :executable ) { <<-RUBY.gsub(/^ */, "").strip }
#{shebang}
require " rack "
puts " EXEC: \# {caller.grep(/load/).empty? ? 'exec' : 'load'} "
puts " ARGS: \# {$0} \# {ARGV.join(' ')} "
puts " RACK: \# {RACK} "
process_title = ` ps -o args -p \# {Process.pid} ` . split ( " \n " , 2 ) . last . strip
puts " PROCESS: \# {process_title} "
RUBY
before do
path . open ( " w " ) { | f | f << executable }
path . chmod ( 0 o755 )
install_gemfile <<-G
gem " rack "
G
end
let ( :exec ) { " EXEC: load " }
let ( :args ) { " ARGS: #{ path } arg1 arg2 " }
let ( :rack ) { " RACK: 1.0.0 " }
let ( :process ) do
title = " PROCESS: #{ path } "
2019-04-14 02:01:35 -04:00
title += " arg1 arg2 "
2018-11-02 19:07:56 -04:00
title
end
let ( :exit_code ) { 0 }
let ( :expected ) { [ exec , args , rack , process ] . join ( " \n " ) }
let ( :expected_err ) { " " }
2019-04-14 02:01:35 -04:00
subject { bundle " exec #{ path } arg1 arg2 " }
2018-11-02 19:07:56 -04:00
shared_examples_for " it runs " do
it " like a normally executed executable " do
subject
expect ( exitstatus ) . to eq ( exit_code ) if exitstatus
2019-06-01 05:49:40 -04:00
expect ( err ) . to eq ( expected_err )
expect ( out ) . to eq ( expected )
2018-11-02 19:07:56 -04:00
end
end
it_behaves_like " it runs "
context " the executable exits explicitly " do
let ( :executable ) { super ( ) << " \n exit #{ exit_code } \n puts 'POST_EXIT' \n " }
context " with exit 0 " do
it_behaves_like " it runs "
end
context " with exit 99 " do
let ( :exit_code ) { 99 }
it_behaves_like " it runs "
end
end
context " the executable exits by SignalException " do
let ( :executable ) do
ex = super ( )
ex << " \n "
ex << " raise SignalException, 'SIGTERM' \n "
ex
end
2019-04-14 02:01:35 -04:00
let ( :expected_err ) { " " }
2018-11-02 19:07:56 -04:00
let ( :exit_code ) do
# signal mask 128 + plus signal 15 -> TERM
# this is specified by C99
128 + 15
end
it_behaves_like " it runs "
end
2019-01-04 08:10:58 -05:00
context " the executable is empty " do
2018-11-02 19:07:56 -04:00
let ( :executable ) { " " }
let ( :exit_code ) { 0 }
let ( :expected_err ) { " #{ path } is empty " }
let ( :expected ) { " " }
it_behaves_like " it runs "
end
2019-01-04 08:10:58 -05:00
context " the executable raises " do
2018-11-02 19:07:56 -04:00
let ( :executable ) { super ( ) << " \n raise 'ERROR' " }
let ( :exit_code ) { 1 }
let ( :expected_err ) do
" bundler: failed to load command: #{ path } ( #{ path } ) " \
" \n RuntimeError: ERROR \n #{ path } :10:in `<top (required)>' "
end
it_behaves_like " it runs "
end
2019-01-04 08:10:58 -05:00
context " the executable raises an error without a backtrace " do
2018-11-02 19:07:56 -04:00
let ( :executable ) { super ( ) << " \n class Err < Exception \n def backtrace; end; \n end \n raise Err " }
let ( :exit_code ) { 1 }
let ( :expected_err ) { " bundler: failed to load command: #{ path } ( #{ path } ) \n Err: Err " }
let ( :expected ) { super ( ) }
it_behaves_like " it runs "
end
Fix some bundler specs (#2380)
* These seem to consistenly pass already
* Show actual command when running `make test-bundler`
Current the setup command that installs the necessary gems for testing
bundler was printed, but not the actual command that runs the tests.
That was a bit confusing.
* Borrow trick from setproctitle specs
* A title that long doesn't get set sometimes
No idea why, but the test doesn't need that the title is that long.
* Fix most gem helper spec ruby-core failures
* Fix the rest of the gem helper failures
* Fix version spec by improving the assertion
* Remove unnecessary `BUNDLE_RUBY` environment var
We can use `RUBY` when necessary, and `BUNDLE_RUBY` is not a good name
because bundler considers `BUNDLE_*` variables as settings.
* Rename `BUNDLE_GEM` to `GEM_COMMAND`
This is more descriptive I think, and also friendlier for bundler
because `BUNDLE_` env variables are interpreted by bundler as settings,
and this is not a bundler setting.
This fixes one bundler spec failure in config specs against ruby-core.
* Fix quality spec when run in core
Use the proper path helper.
* Fix dummy lib builder to never load default gems
If a dummy library is named as a default gem, when requiring the library
from its executable, the default gem would be loaded when running from
core, because in core all default gems share path with bundler, and thus
they are always in the $LOAD_PATH. We fix the issue by loading lib
relatively inside dummy lib executables.
* More exact assertions
Sometimes I have the problem that I do some "print debugging" inside
specs, and suddently the spec passes. This happens when the assertion is
too relaxed, and the things I print make it match, specially when they
are simple strings like "1.0" than can be easily be part of gem paths
that I print for debugging.
I fix this by making a more exact assertion.
* Detect the correct shebang when ENV["RUBY"] is set
* Relax assertion
So that the spec passes even if another paths containing "ext" are in
the load path. This works to fix a ruby-core issue, but it's a better
assertion in general. We just want to know that the extension path was
added.
* Use folder structure independent path helper
It should fix this spec for ruby-core.
* Fix the last failing spec on ruby-core
* Skip `bundle open <default_gem>` spec when no default gems
2019-08-19 20:46:31 -04:00
context " when the file uses the current ruby shebang " do
2018-11-02 19:07:56 -04:00
let ( :shebang ) { " # ! #{ Gem . ruby } " }
it_behaves_like " it runs "
end
2019-01-04 08:10:58 -05:00
context " when Bundler.setup fails " , :bundler = > " < 3 " do
2018-11-02 19:07:56 -04:00
before do
gemfile <<-G
gem 'rack' , '2'
G
ENV [ " BUNDLER_FORCE_TTY " ] = " true "
end
let ( :exit_code ) { Bundler :: GemNotFound . new . status_code }
2019-04-14 02:01:35 -04:00
let ( :expected ) { " " }
let ( :expected_err ) { <<-EOS.strip }
2018-11-02 19:07:56 -04:00
\ e [ 31 mCould not find gem 'rack (= 2)' in any of the gem sources listed in your Gemfile . \ e [ 0 m
\ e [ 33 mRun ` bundle install ` to install missing gems . \ e [ 0 m
EOS
it_behaves_like " it runs "
end
2019-01-04 08:10:58 -05:00
context " when Bundler.setup fails " , :bundler = > " 3 " do
2018-11-02 19:07:56 -04:00
before do
gemfile <<-G
gem 'rack' , '2'
G
ENV [ " BUNDLER_FORCE_TTY " ] = " true "
end
let ( :exit_code ) { Bundler :: GemNotFound . new . status_code }
2019-04-14 02:01:35 -04:00
let ( :expected ) { " " }
let ( :expected_err ) { <<-EOS.strip }
2018-11-02 19:07:56 -04:00
\ e [ 31 mCould not find gem 'rack (= 2)' in locally installed gems .
The source contains 'rack' at : 1 . 0 . 0 \ e [ 0 m
\ e [ 33 mRun ` bundle install ` to install missing gems . \ e [ 0 m
EOS
it_behaves_like " it runs "
end
context " when the executable exits non-zero via at_exit " do
let ( :executable ) { super ( ) + " \n \n at_exit { $! ? raise($!) : exit(1) } " }
let ( :exit_code ) { 1 }
it_behaves_like " it runs "
end
context " when disable_exec_load is set " do
let ( :exec ) { " EXEC: exec " }
let ( :process ) { " PROCESS: ruby #{ path } arg1 arg2 " }
before do
2019-04-14 02:01:35 -04:00
bundle " config set disable_exec_load true "
2018-11-02 19:07:56 -04:00
end
it_behaves_like " it runs "
end
context " regarding $0 and __FILE__ " do
let ( :executable ) { super ( ) + <<-'RUBY' }
puts " $0: #{ $0 . inspect } "
puts " __FILE__: #{ __FILE__ . inspect } "
RUBY
let ( :expected ) { super ( ) + <<-EOS.chomp }
$0 : #{path.to_s.inspect}
__FILE__ : #{path.to_s.inspect}
EOS
it_behaves_like " it runs "
context " when the path is relative " do
let ( :path ) { super ( ) . relative_path_from ( bundled_app ) }
2019-04-14 02:01:35 -04:00
it_behaves_like " it runs "
2018-11-02 19:07:56 -04:00
end
context " when the path is relative with a leading ./ " do
let ( :path ) { Pathname . new ( " ./ #{ super ( ) . relative_path_from ( Pathname . pwd ) } " ) }
2019-04-14 02:01:35 -04:00
pending " relative paths with ./ have absolute __FILE__ "
2018-11-02 19:07:56 -04:00
end
end
context " signal handling " do
let ( :test_signals ) do
open3_reserved_signals = %w[ CHLD CLD PIPE ]
reserved_signals = %w[ SEGV BUS ILL FPE VTALRM KILL STOP EXIT ]
bundler_signals = %w[ INT ]
Signal . list . keys - ( bundler_signals + reserved_signals + open3_reserved_signals )
end
context " signals being trapped by bundler " do
let ( :executable ) { strip_whitespace <<-RUBY }
#{shebang}
begin
Thread . new do
puts 'Started' # For process sync
STDOUT . flush
sleep 1 # ignore quality_spec
raise " Didn't receive INT at all "
end . join
rescue Interrupt
puts " foo "
end
RUBY
2019-04-14 02:01:35 -04:00
it " receives the signal " do
2018-11-02 19:07:56 -04:00
bundle! ( " exec #{ path } " ) do | _ , o , thr |
o . gets # Consumes 'Started' and ensures that thread has started
Process . kill ( " INT " , thr . pid )
end
expect ( out ) . to eq ( " foo " )
end
end
context " signals not being trapped by bunder " do
let ( :executable ) { strip_whitespace <<-RUBY }
#{shebang}
signals = #{test_signals.inspect}
result = signals . map do | sig |
Signal . trap ( sig , " IGNORE " )
end
puts result . select { | ret | ret == " IGNORE " } . count
RUBY
it " makes sure no unexpected signals are restored to DEFAULT " do
test_signals . each do | n |
Signal . trap ( n , " IGNORE " )
end
bundle! ( " exec #{ path } " )
expect ( out ) . to eq ( test_signals . count . to_s )
end
end
end
end
context " nested bundle exec " do
2019-05-24 11:46:26 -04:00
context " when bundle in a local path " do
2018-11-02 19:07:56 -04:00
before do
gemfile <<-G
2019-05-06 12:06:21 -04:00
source " #{ file_uri_for ( gem_repo1 ) } "
2018-11-02 19:07:56 -04:00
gem " rack "
G
2019-08-16 11:51:18 -04:00
bundle " config set path vendor/bundler "
[bundler/bundler] Don't use system bundler on this spec
If we use system bundler, when booting the "outermost" bundler process,
bundler will save the path to the system bundler in BUNDLE_BIN_PATH, and
use it again when booting the "innermost" bundler process (`bundle exec
echo foo`).
That means that second process will use the system bundler path again.
However, we have `-rsupport/hax` in RUBYOPT, so that file will load from
the local copy of bundler, and that file will load `bundler/version`
from the project (not from system), because -Ilib is in the LOAD_PATH.
That will end up causing redefinition errors because the same constant
will be loaded from two different locations.
In general, this is expected behavior, normally you will wrap the
process with `Bundler.with_original_env` to reset the environment.
However, the easiest fix here is to not use system bundler, because it's
not really necessary and thus doesn't help the readability of the spec.
https://github.com/bundler/bundler/commit/a3d72a34ab
2019-08-16 11:54:12 -04:00
bundle! :install
2018-11-02 19:07:56 -04:00
end
2019-05-24 11:46:26 -04:00
it " correctly shells out " , :ruby_repo do
2018-11-02 19:07:56 -04:00
file = bundled_app ( " file_that_bundle_execs.rb " )
create_file ( file , <<-RB)
#!#{Gem.ruby}
2019-07-26 07:37:51 -04:00
puts ` bundle exec echo foo `
2018-11-02 19:07:56 -04:00
RB
file . chmod ( 0 o777 )
[bundler/bundler] Don't use system bundler on this spec
If we use system bundler, when booting the "outermost" bundler process,
bundler will save the path to the system bundler in BUNDLE_BIN_PATH, and
use it again when booting the "innermost" bundler process (`bundle exec
echo foo`).
That means that second process will use the system bundler path again.
However, we have `-rsupport/hax` in RUBYOPT, so that file will load from
the local copy of bundler, and that file will load `bundler/version`
from the project (not from system), because -Ilib is in the LOAD_PATH.
That will end up causing redefinition errors because the same constant
will be loaded from two different locations.
In general, this is expected behavior, normally you will wrap the
process with `Bundler.with_original_env` to reset the environment.
However, the easiest fix here is to not use system bundler, because it's
not really necessary and thus doesn't help the readability of the spec.
https://github.com/bundler/bundler/commit/a3d72a34ab
2019-08-16 11:54:12 -04:00
bundle! " exec #{ file } "
2018-11-02 19:07:56 -04:00
expect ( out ) . to eq ( " foo " )
end
end
context " with a system gem that shadows a default gem " do
let ( :openssl_version ) { " 99.9.9 " }
let ( :expected ) { ruby " gem 'openssl', '< 999999'; require 'openssl'; puts OpenSSL::VERSION " , :artifice = > nil }
it " only leaves the default gem in the stdlib available " do
skip " openssl isn't a default gem " if expected . empty?
install_gemfile! " " # must happen before installing the broken system gem
build_repo4 do
build_gem " openssl " , openssl_version do | s |
s . write ( " lib/openssl.rb " , <<-RB)
raise " custom openssl should not be loaded, it's not in the gemfile! "
RB
end
end
system_gems ( :bundler , " openssl- #{ openssl_version } " , :gem_repo = > gem_repo4 )
file = bundled_app ( " require_openssl.rb " )
create_file ( file , <<-RB)
#!/usr/bin/env ruby
require " openssl "
puts OpenSSL :: VERSION
warn Gem . loaded_specs . values . map ( & :full_name )
RB
file . chmod ( 0 o777 )
aggregate_failures do
expect ( bundle! ( " exec #{ file } " , :artifice = > nil ) ) . to eq ( expected )
expect ( bundle! ( " exec bundle exec #{ file } " , :artifice = > nil ) ) . to eq ( expected )
expect ( bundle! ( " exec ruby #{ file } " , :artifice = > nil ) ) . to eq ( expected )
2019-04-14 02:01:35 -04:00
expect ( run! ( file . read , :artifice = > nil ) ) . to eq ( expected )
2018-11-02 19:07:56 -04:00
end
# sanity check that we get the newer, custom version without bundler
sys_exec ( " #{ Gem . ruby } #{ file } " )
2019-06-01 05:49:40 -04:00
expect ( err ) . to include ( " custom openssl should not be loaded " )
2018-11-02 19:07:56 -04:00
end
end
end
end