Regression introduced in v0.19.0 to config.toml

[walpox]

Bug Report

Environment

Zola version: 0.19.1

Expected Behavior

A site which was built fine with version 0.17.2 of Zola should also be built fine with version 0.19.1, considering no changes to its files, neither to its config.toml file.

Current Behavior

The Zola site, which was built with version 0.17.2, has two relevant keys defined in its config.toml file:

1. generate_feed
2. feed_filename

When it is attempted to build this site with version 0.19.1 of Zola, parsing errors for config.toml are thrown for both of those keys. It seems that the nouns of those keys now need to be written in their plural forms and that the feed_filename(s) key seems to have become a string array, rather than a single string.

Step to reproduce

1. Build a site with version 0.17.2 of Zola, ensuring the generate_feed and the feed_filename keys are set explicitly to valid values.
2. Check that the site builds without errors (e.g. with zola serve).
3. Build the site again, this time with version 0.19.1 of Zola.
4. The zola serve command should throw TOML parsing errors now for config.toml.

Conclusion

If Zola follows semantic versioning, wouldn't it be reasonable to consider this change to break backwards compatibility?

Thus, either implement that functionality in a way which supports backwards compatibility, or postpone it for a version 1.x.y of Zola.

I don't know much about Rust or coding yet, but I think the change was implemented in this PR:

#2477

The problem to consider here is environments where Zola is set to auto-update on its minor and patch numbers, ignoring the major number. For example: "0.x.y".

If Zola turns out not to follow semantic versioning and this is expected behavior, my apologies in advance.

[Keats]

If Zola follows semantic versioning, wouldn't it be reasonable to consider this change to break backwards compatibility?

It does follow semantic versioning, and you can have breaking changes in any versions before 1.0 (https://semver.org/#spec-item-4). That change bumped the y version to indicate the breaking changes.
You're right though that if we were after 1.0.0, we would have had to bump to 2.0.0

[walpox]

Oh, I didn't know about point of the semantic versioning spec. Thank you for clarifying.