[rustdoc] Correctly handle output options with --show-coverage

[GuillaumeGomez]

Fixes #158929.

The -o option was ignored with --show-coverage. This PR takes it into account and makes it closer to how these options are handled in the rest of rustdoc.

Considering calculate_doc_coverage was never really a pass and that I needed RenderOptions (for the -o option), I moved it out of passes.

r? @camelid

[rustbot]

camelid is currently at their maximum review capacity.
They may take a while to respond.

[GuillaumeGomez]

Ah, maybe @Urgau then?

r? @Urgau

[Urgau]

I haven't looked at the implementation, but it seems a bit weird to me to have an flag called --show-coverage that doesn't show anything when used with -o. Are we worried about confusing users?

#158929 suggests updating the documentation instead.
cc @aDotInTheVoid (in case you have an opinion)

[GuillaumeGomez]

I haven't looked at the implementation, but it seems a bit weird to me to have an flag called --show-coverage that doesn't show anything when used with -o. Are we worried about confusing users?

#158929 suggests updating the documentation instead. cc @aDotInTheVoid (in case you have an opinion)

If you want to generate the output in a file, that seems kinda logical to not have output on stdout.

[Urgau]

If you want to generate the output in a file, that seems kinda logical to not have output on stdout.

100% agree, but the CLI option is called --show-..., that implies (at least to me) -o - (i.e. the current behavior). Maybe we should have --coverage?

Is there a way to know if some users are currently relying on the current behavior? (since this is a breaking change)

[GuillaumeGomez]

docs.rs. We could rename the option, I don't mind. But something like --doc-coverage seems better than just coverage, no?

[Urgau]

Yeah, --doc-coverage seems fine. I'm mostly worried about the breaking change to --show-coverage, so if -o is respected only for this new option, that seems fine to me.