Ruby buildpack failures after updating to workflow 2.13.0

[nathansamson]

This weekend we updated to workflow 2.13, mainly to fix the problem of a crashing builder every so often.

But this had the unintended side effect that deploying our rails app did not work anymore as it seems to ignore environment variables during building.
Let me explain:

We need the CURL_TIMEOUT environment variable to be set, as for some reason our european cluster has a very bad download time to Amazon S3 overseas, and will always faill with the default of 30s.
This should still work according to: https://github.com/heroku/heroku-buildpack-ruby/blob/master/lib/language_pack/fetcher.rb#L38

The only solution to get it working for is to manually set the BUILD_PACKURL to v149 (very well possible that v150 for example might also work, but I believe v149 is the one we used before the workflow update).

I can't really imagine this is a bug in deis builder as clearly the v149 buildpack respects my settings, but on the other hand I am not sure why the env variable wouldn't work anymore in the newer buildback...

Any ideas/suggestions?

[mboersma]

@nathansamson slugbuilder is now up to ruby buildpack v159 in master. I don't see a smoking gun in the diff between v149 and v159 though. Could you try v159 so we know for sure?

[szymonpk]

@nathansamson Have you managed to fix this, differently than forking buildpack? Where have you tried to set CURL_TIMEOUT? In builder deployment?

[Cryptophobia]

This issue was moved to teamhephy/builder#11