Open beruic opened 3 years ago
Beat me to it! I'm seeing this manifested in a huge swath of the google-cloud-ruby libraries, particularly google cloud storage.
Since the issue is experiences in fastlane
with uploading app packages to Google Play, I guess there might be a Google angle to it.
Without knowing to much about Ruby at all, I'm guessing that either package just introduced a breaking change or a bug, or this package fixed something that has been misused by despendants.
Perhaps the representable
package should be made aware of this, since that seems to be what we have in common here.
I'm happy to open an issue on that repo if you aren't already doing so.
Done
This regression seems to be because of https://github.com/apotonick/declarative-option/pull/4. In particular, this code is a breaking change:
# code changed from:
->(context, *args) { context.instance_exec(*args, &value) }
# to:
->(context, *args, keyword_arguments: {}, **) do
context.instance_exec(*args, **keyword_arguments, &value)
end
The second form is not equivalent to the first, the **
will capture and discard keyword arguments passed in.
Should this release be updated to the MINOR (even MAJOR) version instead of PATCH?
Yes, we already yanked the faulty 0.1.1 version and will prepare everything so that Ruby-3 gems use the correct version. Most likely, we will drop this gem and instead use trailblazer-option
, since we use the Option()
concept almost everywhere. Don't worry, this won't bring any additional dependency. :beers:
Well, this is really misunderstanding. Feels like the left-pad issue: https://qz.com/646467/how-one-programmer-broke-the-internet-by-deleting-a-tiny-piece-of-code/
I'd say we shouldn't do this in the future. RubyGems should not allow this either.
@imWildCat You mean RubyGems shouldn't allow yanking? I am glad it exists - I wouldn't have been able to provide a real fix quickly. I am in the process of giving over control of several gems to my fellow OSS devs and as you can see, the process is not fully QAed, yet :stuck_out_tongue_winking_eye: Sorry for the hassle!
Oh, that's an interesting article you linked, thanks a lot! :beers:
I respect your motivation but it might be better to bump the version to apply a fix.
I spent about 1 hour today to investigate this bug. For some reason, Google didn't show me this GH issue at the very beginning.
I feel that in this case yanking was appropriate: it clearly communicates that the faulty version does not exist anymore, so one can check why, and what versions are available instead. Otherwise, finding this issue would be even harder.
Todays update seems to have a bug that manifests itself in
fastlane
. Please see my comment at https://github.com/fastlane/fastlane/issues/18403#issuecomment-800657944The stack trace I get is: