Set "Accept-Encoding: identity" header when curl'ing sources

[vrothberg]

Bringing the conversation from https://gitlab.com/redhat/hummingbird/rpms/-/merge_requests/66#note_2984270969 over here:

Looks like your tools are affected by https://pagure.io/fedora-infrastructure/issue/12812 or a similar issue from other servers that suffer from "content encoding confusion".
You will need to set the Accept-Encoding: identity HTTP header when doing source downloads, otherwise you get an on-the-fly un-gzipped file with a ".tar.gz" file extension (that is actually just an uncompressed "tar" file).

Setting header = \"Accept-Encoding: identity\" in the .curlrc fixes the issue but I think it would be great if dist-git could take care of it directly.

[xsuchy]

The only thing is that we set on the server content-type: application/octet-stream, but I believe that is wrong. The content is gzip-ed. And when we know, we should indicate it. It is client responsibility to indicate what he wants. And act accordingly.

[xsuchy]

Hmm, we can add it here https://github.com/release-engineering/dist-git/blob/main/dist-git-client/dist_git_client.py#L111

[vrothberg]

Hmm, we can add it here https://github.com/release-engineering/dist-git/blob/main/dist-git-client/dist_git_client.py#L111

Yes, that's what I had in mind. As an immediate patch, we use the .curlrc trick (see https://gitlab.com/redhat/hummingbird/rpms/-/merge_requests/144/diffs?commit_id=adc35f627bb9c9bc948e67a332bf15a1fc6fdad6) but other users may hit it as well.