Bitbucket Cloud: multi-line inline comments collapse onto their first line (inline.start_to is never sent)

[bannert1337]

What happens

A review finding that spans several lines is posted on Bitbucket Cloud as a single-line comment anchored to the first line of the span. The rest of the range is dropped, so a finding about a whole function reads as a remark about its signature.

Why

BitbucketProvider.publish_inline_comments builds the payload as

"inline": {"path": ..., "to": position}

using only the start line. Bitbucket Cloud's comment API does accept a range: inline.start_to together with inline.to. The API validates it ("start_to" must be less than or equal to "to"; an inverted range is rejected with 400) and stores it - a GET on the created comment returns both fields.

The community post commonly cited for "Bitbucket does not support ranges" is about from/to, which are the old-side and new-side line numbers, not a span.

Suggested fix

Send the range when the finding has one: inline.start_to = relevant_lines_start, inline.to = relevant_lines_end, falling back to the single-line form when they are equal or the end line is missing.

Verified against a live Bitbucket Cloud pull request on 2026-09-10: a comment published with relevant_lines_start=10, relevant_lines_end=16 comes back from the API as {"path": ..., "start_to": 10, "to": 16} and renders spanning those lines.

We carry this as a patch on v0.45.0 and are happy to open a pull request.

[IsmaelMartinez]

Thanks for going back to the API to check what start_to actually does, and for catching that the cited community post is about from/to.

Confirmed. A 10 to 16 span through publish_inline_comments sends {"inline": {"to": 10, "path": ...}}, so the range is gone.

Higher priority than it looks. is_supported("gfm_markdown") is False for Bitbucket Cloud, so the table branch at pr_code_suggestions.py:314 never runs and stock /improve takes the inline path, and the bitbucket_app webhook default passes commitable_code_suggestions=true. This is the default path, not an opt-in one.

Please open it scoped to BitbucketProvider, with a test asserting the payload, and delete the stale comment at line 504.