We've been experiencing lately some transient failures with rspec tests on Windows GitHub Actions. It takes about 16 minutes to fail, whereas successful runs happen in about 18 minutes, so this happens very late and can easily become annoying.
Over 900 tests (958 seen so far) across multiple files fail when this happen (from a total of 26109 examples). Most of them are due to objects being nil and methods cannot be called (around 940).
We've been experiencing lately some transient failures with [rspec tests|https://github.com/puppetlabs/puppet/actions/workflows/rspec_tests.yaml] on Windows GitHub Actions. It takes about 16 minutes to fail, whereas successful runs happen in about 18 minutes, so this happens very late and can easily become annoying.
Over 900 tests (958 seen so far) across multiple files fail when this happen (from a total of 26109 examples). Most of them are due to objects being *nil* and methods cannot be called (around 940).
Attached logs are from are from such failed run (see *windows-2016(ruby 2.6)* files).
2021-07-21T07:26:16.7344142Z 1) Puppet::Util::Windows::Process when reading environment variables will ignore only keys or values with corrupt byte sequences
2021-07-21T07:26:16.7346563Z Memory access offset=32 size=4 is out of bounds
which occurs in an after each block. That leaves puppet's context in a bad state causing the rest of the failures. I'm going to close this as a dup of PUP-11122