Similar to docker-library/golang#552, there are a lot of upstream tests we could run and should run somewhere, but during build doesn't necessarily make sense.
In https://salsa.debian.org/ruby-team/ruby/-/blob/5b91c2c6b64195da7110d341daba904b9d12c9d5/debian/tests/run-all, we can see that Debian does some clever things to run a lot of the upstream tests post-build via "autopkgtests" (which is very akin to how we run tests on our images after they're built -- they typically run against the built + installed packages).
The basic gist is ruby test/runner.rb, and Debian's got a pretty hefty collection of tests they exclude which will probably inspire tests we need to exclude too (especially # FIXME for now, we are excluding the tests for C extensions; couldn't figure out how to properly build them without building everything else, since we'll have the same issue with doing this post-build).
So we'll need to recreate our "download the source code" logic to get the test files back, but that's the easy part. My own very basic testing of this idea on ruby:slim comes up with "7 failures, 6 errors, 176 skips", so there's clearly some work to do here (for each of those, fixing the image, accommodating the test somehow with runtime flags/packages/etc, or building up a list of tests to skip and writing down justifications for why we think they're valid to skip/ignore).
Similar to docker-library/golang#552, there are a lot of upstream tests we could run and should run somewhere, but during build doesn't necessarily make sense.
In https://salsa.debian.org/ruby-team/ruby/-/blob/5b91c2c6b64195da7110d341daba904b9d12c9d5/debian/tests/run-all, we can see that Debian does some clever things to run a lot of the upstream tests post-build via "autopkgtests" (which is very akin to how we run tests on our images after they're built -- they typically run against the built + installed packages).
The basic gist is
ruby test/runner.rb, and Debian's got a pretty hefty collection of tests they exclude which will probably inspire tests we need to exclude too (especially# FIXME for now, we are excluding the tests for C extensions; couldn't figure out how to properly build them without building everything else, since we'll have the same issue with doing this post-build).So we'll need to recreate our "download the source code" logic to get the test files back, but that's the easy part. My own very basic testing of this idea on
ruby:slimcomes up with "7 failures, 6 errors, 176 skips", so there's clearly some work to do here (for each of those, fixing the image, accommodating the test somehow with runtime flags/packages/etc, or building up a list of tests to skip and writing down justifications for why we think they're valid to skip/ignore).