In the past, I've seen several times where the bundler default gem has caused issues in external CI. It is currently causing issues with CI's that use nested 'bundle' commands.
External repos should not have to force an install/update of RubyGems/Bundler for ruby-head to work.
Also, the lib/bundler/build_metadata.rb file contains no information about the files currently installed in master, and hence neither the gemspec or commands like
bundle version are correct.
There has also been a gem release of bundler that occurred almost two weeks after the last update here. Technically, that version is 2.0.2, while master is using bundler master/2.1.0.pre.1, but I'm sure some of the commits overlap.
Hence, can bundler be updated more frequently and accurately?
Updated by hsbt (Hiroshi SHIBATA) 12 months ago
- Assignee set to hsbt (Hiroshi SHIBATA)
- Status changed from Open to Feedback
I didn't understand what you requested.
If your concern is `build_metadata, I will set it from the upstream commit.
PS. Please set the informational subject always.
Updated by MSP-Greg (Greg L) 12 months ago
Sorry, bad week, and really hot where I am.
Main thing, I would like ruby-head/master/trunk to not break external CI. In some instances, it is currently doing so due to an issue with Bundler. I am also willing to help with that breakage...
Regarding Bundler, I was looking at ruby-loco and a recent Travis build, and
build_metadata has no date or sha, and the gemspec has no date. So, without looking at the git repo, there's no way to determine what's being used, other than a somewhat meaningless version.
I know that requires modifying files, but it would be really helpful to have that info, especially given that RubyGems and Bundler have a constant flow of commits, unlike other std-lib/default gems.
Many people contribute here, but you have actual things/tasks that you apear to be responsible for. I don't want to add to that, but more timely updates to RG/Bundler would be helpful.
I've gotten the 'you have to commit/PR to upstream'. Given that RG/Bundler are dynamic and they can break trunk, maybe the rule for RG/Bundler should be commit/PR in upstream, or both, but not just in trunk...