Closed jordan-brough closed 7 months ago
FYI I tried to repro the failure from this one by running the following locally, but I wasn't able to reproduce the error:
rm -r gemfiles/*.lock
export RBENV_VERSION=2.7.5
bin/setup
bundle exec rspec --seed 3813
bundle exec appraisal install
BUNDLE_GEMFILE=gemfiles/rails50.gemfile bundle exec rspec --seed 38463
BUNDLE_GEMFILE=gemfiles/rails51.gemfile bundle exec rspec --seed 18633
BUNDLE_GEMFILE=gemfiles/rails52.gemfile bundle exec rspec --seed 3575
BUNDLE_GEMFILE=gemfiles/rails60.gemfile bundle exec rspec --seed 42679 # This one failed in Circle with a segfault
If this is something that has occasionally happened in the past, then it could be that some of the extra flakiness is due to the fact that we're now running more tests on Circle, due to #2125 being merged?
I verified that for the example in my comment above (the "FYI I tried to repro the failure ..." one) that the test that segfaulted was the first feature
test run for that commit + seed
value. And the stack trace shows it's executing some sprockets
and concurrent-ruby
code.
I wonder if any of the comments in https://github.com/sass/sassc-ruby/issues/207 are applicable?
Hmm, interesting. I've seen this in the past and it seems much worse now.
I just pushed up this branch with a potential solution. I'm going to run it a few times and see how it behaves on CI: https://github.com/thoughtbot/administrate/compare/nc-sassc-seg-fault-fix?expand=1
I have an app that I have been to upgrade to cssbundling-rails
. Unfortunately, I have faced the Segmentation fault
errors. After trying everything with no success I want to get rid of the sassc-rails
gem. Is it possible to run administrate without this gem?
That branch did no good and I'm not necessarily convinced that spending much time on it is worthwhile.
@gathuku, hmm, good question. I don't see why you wouldn't in principle be able to do so. But you might need to fork the gem and change the dependencies around so that it's never bundled.
With https://github.com/thoughtbot/administrate/pull/2397 merged, we are not using sassc any more. I think this should be fixed...?
If not let us know and we'll reopen.
The Ruby 2.6 & 2.7 tests seem to have suddenly become flaky:
I saw some failures like that in a couple of my PRs, that usually went away if I retried the specs (e.g. by pushing an empty commit).
It might be something from one of my recent commits? It doesn't look like they were flaky before that, unless y'all have already been retrying failures like that on master branch when they happened?
Example failure: https://app.circleci.com/pipelines/github/thoughtbot/administrate/1662/workflows/238e328c-cdd2-4f65-ae6a-24f9789c20c4/jobs/8803
Some example outputs: