The scraper http client adapter was changed since our last run, and the newer adapter requires the HTTP_PROXY and HTTPS_PROXY variables to be manually wired into the adapter proxy options. This is necessary on the analytics clients where outbound traffic must go through a proxy, see /etc/profile.d/proxy.sh.

Options should be in the form given in https://hexdocs.pm/mint/Mint.HTTP.html#connect/4-proxying

Hardcoding the options would look like this:

Wiki.Enterprise.new(username: username(), refresh_token: refresh_token(), connect_options: [proxy: {:http, "webproxy", 8080, []}])

However, the mediawiki-client library should pull proxy configuration from HTTP_PROXY, HTTPS_PROXY, http_proxy, and https_proxy environment variables as a best practice, using System.get_env/1 and falling back through the possible set- or unset-ness.

Code for review:

• https://gitlab.com/adamwight/req_proxy/-/merge_requests/1
• https://gitlab.com/wmde/technical-wishes/mediawiki_client_ex/-/merge_requests/26

See also these references and discussions:

• https://wikitech.wikimedia.org/wiki/HTTP_proxy
• https://curl.se/docs/manpage.html#ENVIRONMENT
• https://about.gitlab.com/blog/we-need-to-talk-no-proxy/
• https://github.com/wojtekmach/req/issues/522