rest2html: treat referenced wrapped images in base document as inlined

[jdknight]

docutils will only add newlines around images it believes are inlined. For images held in references, it checks the parent of the reference if its a TextElement to consider it inlined. Since a document is not a TextElement type, it will wrap an image with newlines. For GitHub output, this is not desired and will result in the extra whitespace being rendered with a reference's decorative line. To avoid this, always strip any appended suffixes for images in this scenario.

---

There are a various GitHub projects which reveal the issue. For example, Sphinx's README.rst shows this issue:

With the changes made in this merge request, the following shows a rendering of HTML before and after the change:

[github-actions]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[jdknight]

!unstale

[github-actions]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[jdknight]

!unstale

[github-actions]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[jdknight]

!unstale

[github-actions]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[jdknight]

!unstale

[github-actions]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[jdknight]

!unstale

[Copilot]

Pull request overview

This PR updates the ReST-to-HTML translation to avoid emitting trailing whitespace after an image when the image is wrapped in a reference that is a direct child of the document, preventing GitHub’s link decoration from rendering across the whitespace after the image.

Changes:

• Strip the trailing newline emitted for a referenced top-level image so it behaves like an inline image in GitHub HTML output.

Show a summary per file

File	Description
lib/github/commands/rest2html	Adjusts depart_image output to remove trailing newline for document-level referenced images to avoid decorative underline/whitespace artifacts.

Copilot's findings

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

• Files reviewed: 1/1 changed files
• Comments generated: 1

[Copilot]

This change adjusts HTML whitespace semantics in a subtle way, but the existing markup tests normalize/strip blank text nodes via Nokogiri (noblanks), so they likely won’t fail if the newline/whitespace regresses. Consider adding a targeted regression test that asserts the raw HTML output for a top-level .. image:: with :target: does not contain whitespace between the <img> and the closing </a> (string-level assertion, not DOM-equality).

[zkoppert]

Hey @jdknight — thanks for sticking with this PR, and sorry it took so long to get eyes on it.

I rebased your branch onto master to get CI running (it was 46 commits behind), and all checks pass across Ruby 3.2/3.3/3.4. The fix itself looks good — we reviewed it thoroughly and the logic is sound.

One thing we'd like to see before merging: a regression test. The existing test suite uses Nokogiri with noblanks, which normalizes whitespace-only text nodes — so the current tests wouldn't catch it if this behavior regressed. A string-level assertion (not DOM-equality) that verifies there's no trailing whitespace between the <img> and closing </a> for a top-level referenced image would do the trick.

Something like an RST fixture with a .. image:: directive using :target:, and an assertion on the raw HTML output before Nokogiri parsing. Happy to help brainstorm the approach if that would be useful.

[jdknight]

@zkoppert, there appears to be a couple of locations in assert_html_equal/normalize_html that would need to be adjusted to allow assert_html_equal to support strict whitespace checks, let alone finding a graceful way to opt-in to strict checks through the generic markup test discovery logic.

Maybe for this case a simple test can be appended. For example:

 def test_blanks_rst_inlined_reference
expected = '<img src="https://example.com/img.svg"></a>'
actual = GitHub::Markup.render_s(GitHub::Markups::MARKUP_RST, <<~RST
.. image:: https://example.com/img.svg
:target: https://example.com/
RST
)
assert_equal expected, actual.strip.lines.last
end

And if more tests are needed later that require strict whitespace checks, this can be looked at then.

[github-actions]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[jdknight]

!unstale