mini_racer: 0.5.0.pre fails to install/build on Github Actions (Ruby 3.0.1)

Not sure this is useful, here what worked for me on Linux:

gem uninstall libv8-node  # to get rid of musl version
bundle lock --add-platform x86_64-linux
gem update --system  # not sure if necessary but I didn't try again without it
bundle update --bundler

This took longer than I’d like to admit; I’m no bundler expert. A combination of @jarthod and @brilnius suggestions fixed this issue for me. My company supports multiple platforms, caches gems locally in vendor/cache, and was using bundler version 1.17.3. Here’s what I did:

1. Uninstall mini_racer via bundle remove mini_racer.
2. Uninstall libv8-node via bundle remove libv8-node if it’s an explicit requirement in your Gemfile. If you have multiple platforms supported in your Gemfile.lock, remove them. There should now be no trace of mini_racer or libv8-node.
3. Ensure all of the supported platforms are listed under PLATFORM in your Gemfile.lock, using bundle lock --add-platform <platform> to add those that may be missing.
4. Update bundler via bundle update --bundler. Your Gemfile.lock should update BUNDLED WITH from 1.17.3 to 2.1.x.
5. Ensure you have bundler 2.3.x (or later) installed via gem install bundler
6. Manually change BUNDLED WITH from 2.1.x to 2.3.x or later (I used the latest available, which was 2.3.4)
7. Install mini_racer via bundle add mini_racer. This will also install its libv8-node dependencies for your local platform.
8. Ensure all platform gems are installed into your local cache via bundle package --all-platforms.

This was all that was required. Sanity check that you’re good to go with a bundle install, which should be a noop.

libv8-node-16.10.0.0-x86_64-linux-musl/vendor/v8/x86_64-linux

Looks like a case of rubygems/rubygems#3174: Ubuntu is glibc so bundler has no business installing a x86_64-linux-musl version, but does and installs it. Then it fails at extension build time because the runtime platform lookup for the static lib at x86_64-linux doesn’t exist within the x86_64-linux-musl gem.

I was updating gem in RoR project and was getting same error. Removing Gemfile.lock and re-installing all gems through

bundle install

resolved this problem. tested on: ruby 2.7.4, RoR 6.1.4.1

Github CI remains unwilling to use 0.5.0 on Ruby 2.6.6, even after deleting Gemfile.lock and updating as many gems as bundler could find. Reverting to 0.4.0 allows CI to start running and our Ubuntu 18 staging server to update gems, so it would seem there continues to be some incompatibility in Ruby 2.6+mini_racer 0.5.0 on Ubuntu-based platforms, which manifests as

compiling mini_racer_extension.cc
cc1plus: warning: command line option ‘-Wimplicit-int’ is valid for C/ObjC but
not for C++
cc1plus: warning: unrecognized command line option ‘-Wno-self-assign’
cc1plus: warning: unrecognized command line option ‘-Wno-parentheses-equality’
cc1plus: warning: unrecognized command line option
‘-Wno-constant-logical-operand’
cc1plus: warning: unrecognized command line option ‘-Wno-cast-function-type’
linking shared-object mini_racer_extension.so
g++: error:
/home/runner/work/gitclear/gitclear/vendor/bundle/ruby/2.6.0/gems/libv8-node-16.10.0.0-x86_64-linux/vendor/v8/x86_64-linux/libv8/obj/libv8_monolith.a:
No such file or directory
Makefile:260: recipe for target 'mini_racer_extension.so' failed
make: *** [mini_racer_extension.so] Error 1

Thanks, for the record I was having the same issue and decided to wait until the issue settles a bit before wasting the same time as everyone else trying the same things. Now I tried again and was able to upgrade “easily” by:

1. Upgrading bundler from 2.1.2 to 2.3.0 (not sure if that’s required in the end?). I used gem install bundler and then changed the version manually in my Gemfile.lock as didn’t find any official way working for me.
2. Adding the x86_64-linux platform to my Gemfile.lock with bundle lock --add-platform x86_64-linux (before it had only ruby and would cause the download of libv8-musl version, now it has both ruby and x86_64-linux and it downloads the proper libv8-linux)

for anyone else unable to get it to work, you need to clobber any installed libv8-node files/directories, as rubygems/bundler can’t remove it properly on it’s own. find those files with:

find / -name '*libv8-node*' 2>/dev/null

this is the only way to get it to work if you’ve attempted to bundle before adding the platform, in my experience.

rails 7 ruby 3.0.4 and 3.1.2 Was running into this when on mac (none M1) build docker-compose, resolved by:

1. from ‘Gemfile.lock’ file remove ‘ruby’ under ‘PLATFORMS’ section

$ gem uninstall libv8-node
$ bundle lock --add-platform x86_64-linux
$ gem update --system
$ bundle update --bundler
$ bundle install

Another workaround that may help people sometimes when you can’t update rubygems/bundler, install libv8-node and mini_racer before bundle install:

# match the versions in your Gemfile.lock!
gem install libv8-node:16.10.0.0 # this should pick up the one for your platform
gem install mini_racer:0.6.2 # this will find the above as a dependency and build using it
bundle install # this may stubbornly install libv8-node for linux-musl again but since mini_racer is already there and built and bundler picks it up that libv8-node is not going to be used

Since you can’t use bundle show/info as bundle install has not been run yet, you can hack around to get locked gem versions automatically:

$ cat Gemfile.lock | ruby -ne 'puts $1 if /^    libv8-node \((.*)\)/'
16.10.0.0-aarch64-linux
16.10.0.0-x86_64-darwin
16.10.0.0-x86_64-linux
$ cat Gemfile.lock | ruby -ne 'puts $1 if /^    mini_racer \((.*)\)/'
0.6.2

How to use that is left as an exercise for the reader.

I think I’m going to close this issue. Maybe we could add a FAQ/remark in the readme about this problem, but adding the “proper” platforms seems to do the trick more or less for everyone.

Also, see this comment I made, which should work around the issue.

I have a second project which had the same problem. Now, that I run bundle update, it simply uses the local mini_racer Using mini_racer 0.5.0 (was 0.4.0) without changing the PLATFORMS section from ruby to whatever. I guess on another machine without having 0.5.0 already installed this will cause the same problem again. For me a real bundler pitfall…

There’s been this security-related change in recent bundler (mentioned above) where you now have to do e.g bundle platform add x86_64-linux, as it’s not automatic anymore to look for additional platforms. This is manual and mandatory to preclude supply-chain attacks. It’s a blocking step only when deployment mode is triggered, but I venture it might still affect non-deployment mode since the lockfile would not have information about the non-added platforms and would attempt automatic stuff that would behave differently than if the platform was filled in.

Also, you can try updating your rubygems+bundler, if that is possible for you.

I find myself with the same problem on kubuntu 21.10. Is there a way to work around it (except using mini_racer 0.4.0) ? Thanks for any advice!