Allow spaces in path to Yarn binstub and only run on precompile if needed #40785
Conversation
One reason why it could be missing because Rails was upgraded but `rails app:update` was not run. Running `rails app:update:bin` should create it. refs rails#40795
| rescue Errno::ENOENT | ||
| $stderr.puts "bin/yarn was not found." | ||
| $stderr.puts "Please run `bundle exec rails app:update:bin` to create it." | ||
| exit 1 |
There was a problem hiding this comment.
Another issue #40795 popped up where apparently bin/yarn not exists. My assumption (to be confirmed) is that rails app:update was not run between Rails upgrades – which should create bin/yarn.
Therefore I added an error message to run rails app:update:bin in such cases where it should create bin/yarn.
One could also remove the exit 1 in such cases so it only prints out the warning but the pipeline continues. It was like this in before Rails 6.1, but I'm not sure if this was the best idea because if Rails requires Yarn to be present it should hard fail if there is something wrong.
doits
changed the title
Based on discussion in rails#40795, it looks like `yarn:install` is *always* run, even if the Rails project disabled javascript and there is no `bin/yarn`. Check for the existence of `bin/yarn` to decide if `yarn:install` should be run or not. The check for this is taken from `railties/lib/rails/app_updater.rb`, where it does the same: ```ruby options[:skip_javascript] = !File.exist?(Rails.root.join("bin", "yarn")) ```
doits
force-pushed
the
yarn_command_with_spaces
branch
from
6a45658
to
87e9ae0
Compare
| end | ||
| end | ||
|
|
||
| # Run Yarn prior to Sprockets assets precompilation, so dependencies are available for use. | ||
| if Rake::Task.task_defined?("assets:precompile") | ||
| if Rake::Task.task_defined?("assets:precompile") && File.exist?(Rails.root.join("bin", "yarn")) |
There was a problem hiding this comment.
After further comments in #40795 it looks like yarn:install is always run, even if javascript was skipped in the Rails project and rails app:update:bin therefore does not generate bin/yarn.
I added this check to make sure bin/yarn exists. It is the same as done in the app updater:
rails/railties/lib/rails/app_updater.rb
Line 24 in b087511
Now it should really only be run automatically before compiling assets if bin/yarn is present.
doits
changed the title
|
@rafaelfranca looks like bd4d8fd made some unintended noise for people updating to Rails 6.1 – I'm sorry for it – but I think I finally found a solution. To summarize: When the task First it was found out that the path to the Yarn binstub was not quoted correctly and failed if it had spaces in it (#40783) and the second issue came up where it runs This PR fixes both issues by quoting the path to the Yarn binstub and by making |
|
What temporary solution is suitable? I really need version 6.1, but because of this problem, I can't update. |
create a #!/bin/bash
exit 0 |
| end | ||
| end | ||
|
|
||
| # Run Yarn prior to Sprockets assets precompilation, so dependencies are available for use. | ||
| if Rake::Task.task_defined?("assets:precompile") | ||
| if Rake::Task.task_defined?("assets:precompile") && File.exist?(Rails.root.join("bin", "yarn")) |
There was a problem hiding this comment.
The only way of this happen is if yarn is not installed and sprockets is, but --skip-javascript doesn't install or even load sprockets. I think this check here is valid though.
There was a problem hiding this comment.
I'm confused. Sprockets is still loaded for CSS, right?
There was a problem hiding this comment.
Right. --skip-javascript is not --skip-assets. I was mixing both
Allow spaces in path to Yarn binstub and only run on precompile if needed
|
Thank you for the fix and merge. |
|
We need this fix to be added to the 6.1.x branch. |
|
@breisig it is there already |
* Turning this into a rack app that you can interact with * More Code Cleanup * WIP * I am going full rails, I can build it from scrach with rack but I just end up building rails * Removed a bunch of ben stubs * Did some extra work got the UI working with webpack * Got the Client to The Movie DB working * More Rubocop cleanup * Added Pagination * Added Rspec * More testing improvements to the STI for config * Added a nice default serializer * Added simple form * More Rubocop cleanup * Added A nice input field for simple form * Got the input for working for config settings to make it easy to add the API key * Added tables to track disk info * Improved the testing coverage * More Features to handle communication with the movie db * Ability to have users with there own configs and preferences * Add Ability to search for movies and TV shows * cleaned up the work a bit * Added the ability to create a movie from the The Movie DB api * Added in wisper to handle callback in a way that is much cleaner * Bundle update * Code Cleanup thanks to rubocop * Added stimulus which and a lot of other fixes * Adding Plex config * Added in plex config * Rubocop fixes * Trying out a remote upload tool * Handle HTTP connections using farday and use FTP to stablish a remote connection * Added TV and better error handling * Added some test to the callbacks when selecting a tv show * Now you can select an season and it will build the episodes * Added a way to now get the episodes * Made everything look a little nicer and added in still pic to episodes * Removed old code that was used to find movie details * Added a way for the system to parse disk titles * Added a smart parser for the MKV output * Cleaned up the code base to make things simple * Rubocop fixes * More testing around MKV parsing * Added new workflow state called selected * Added in the abilty to cancel the work flow * Added abilty to select multiple TV shows on disk * Thank goodness for concurrent-ruby, I can now process the disk info and then notify the FE with websockets * Ability to load disk in the background without slowing down the disk selection process * Removed disk channel that is not being used * Added in a way to rip mkv files as well as a better handling of jobs * Upgrade gems * Fixed up test and make the jobs code a bit cleaner * Rubocop fixes * Annotated Models to make it easier to know what columns exist on a model * Added assocation to movies and disk title id * Improved the workflow so you have to have a disk_title before you can rip * Got the create mkv sort of working * Removed ruby-version file * Fixed borken tests * Update ruby.yml * Run Yarn Install on testing * Added Yarn Install to the workflow * UGH * Created a UI to get info on the worker status * Assocate the Disk Title to the progress so we can store messages on the title * Added messages to titles so we can print some useful info * Trying out caching dependency in github actions * Caching is built into the ruby/setup-ruby * did not need to run bundle handle by ruby setup * Added Rubocop into the mix to see what happens * Cool it has badges nifity * It works damit * Added in some improbments to the boot process * Update Gemfile * Update .tool-versions * code cleanup * stimulus update * stimulus cleanup * Added Stimulus Reflex * Improved the progress code and layouts * Trying to make a better setup script * Building out a MKV downloader to get the latest updates * More Code Cleanup and better disk workers * More Code Cleanup and upgrades * Got the tests passing * Got rubocop all happy, much cleaner * More rubocop cleanup * So much code cleanup needed * Added github actions formater for rspec * Added MKV Installer for Mac * Fixed Yarn Packages use the newest versions * Added a really easy to use setup script * Added enable cache as part of the setup * Don't use rake to enable cacke * More syntax errors thanks to yours * Going super low tech with adding this file * Improved colors and link desc * More database fixes along with a large number of tests * Found more missed files that needed testing * config_make_mkv now points to a stubbed version of makemkvcon * More Tests more code quality improvements * Made the readme a bit nicer and made it test a less * renamed workerflow from ruby to status * Renamed it to status * Improved the validation of the config * Fixed more code issues * Improved DSL for config settings * Movies & TV are the same thing so they got consolidate under Videos * Merged Movies and Tvs into Vidoes table, really they are the same thing with super small diifs * Tried out a new setup script * Added a startup script and install ruby script * Forgot to add fileutils * Not giong to load Rails application * change rake from a command to a another process it just simpler * Make sure we boot the app in production mode * Fixed MKV Downloader * Yarn Install was not able to deal with space but bundle update fixed it rails/rails#40785 * Allow searcing of static assets * Added quick start feature, saving application boot time * Added error handling when a API key is invalid plus only have a single selected title * Rubocop fixes * Added a check to make sure selected does not effect other videos * Broke up the tests a bit to make things a bit easier to read * Added a system to select moves and improved query tools for searching the movie db * Improved Video models and testing around videos * Added caching around the search requests * Got the progress bar working when selecting a movie * Working out some better data tracking for movies * Found some more test I needed to create * Make the progress bar more useful and less generic * Failed to make changed to the view component * Switching over to hotwire turbo * Full VCR update and move some code around to help cleanup the processing * Changed the column time for movie runtime * Removed more complexity with the state tracking * Performance improvements around get request to the movie db * Worked out a way to better cache and store movie titles * More Improvements * Get tests passing again * Run Bundle update * upgraded ruby to 2.7.3 * Upgraded to ruby 2.7.3 and fixed rubocop errors * Guard against nil values when convert_min_to_seconds * Pagination using show hotwire * Reduce unused space * Compact the UI more * Improved Caching and UI to show more data * Reduce the total results per page to just 4 * Made the pagination system more complex. UGH * Got the disk list working again and removed the pagination tools * Huge Upgrade and got progress bar working * Create dependabot.yml * Update dependabot.yml * Disk Space warning along with checks to make sure files get created correctly * Added Encryption for json config settings * Fixed Broken tests and made the configuration tools gooder * Use build in caching to avoid network calls for image data * Cleanup and rspec tests * Got plex ripper working * Got it all working * fixed rubocop
Summary
Fixes #40783, #40795
Before the version check was introduced in bd4d8fd, the
systemcall right below it failed silently in such cases and did nothing. Now the version check call raisesErrno::ENOENTand acutally exposes the missing binary.