Use prefix for TableOfContents filter hrefsjprovazn-ignore-anchors

TableOfContents filter generates hrefs for each header in markdown,
if the header text consists from digits and redacted symbols only,
e.g. "123" or "1.0 then the auto-generated href has the same format
as issue references.

If the generated id contains only digits, then 'anchor-' prefix is
prepended to the id.

Closes #38473

5 files changed, 24 insertions, 1 deletions

diff --git a/app/models/concerns/cache_markdown_field.rb b/app/models/concerns/cache_markdown_field.rb
index 98776eab424..943dc689388 100644
--- a/app/models/concerns/cache_markdown_field.rb
+++ b/app/models/concerns/cache_markdown_field.rb
@@ -11,7 +11,7 @@ module CacheMarkdownField
   extend ActiveSupport::Concern
 
   # Increment this number every time the renderer changes its output
-  CACHE_VERSION = 2
+  CACHE_VERSION = 3
 
   # changes to these attributes cause the cache to be invalidates
   INVALIDATED_BY = %w[author project].freeze
diff --git a/changelogs/unreleased/anchor-issue-references.yml b/changelogs/unreleased/anchor-issue-references.yml
new file mode 100644
index 00000000000..2e60c3a0781
--- /dev/null
+++ b/changelogs/unreleased/anchor-issue-references.yml
@@ -0,0 +1,6 @@
+---
+title: Fix false positive issue references in merge requests caused by header anchor
+  links.
+merge_request:
+author: jprovaznik
+type: fixed
diff --git a/lib/banzai/filter/table_of_contents_filter.rb b/lib/banzai/filter/table_of_contents_filter.rb
index 47151626208..97244159985 100644
--- a/lib/banzai/filter/table_of_contents_filter.rb
+++ b/lib/banzai/filter/table_of_contents_filter.rb
@@ -32,6 +32,7 @@ module Banzai
               .gsub(PUNCTUATION_REGEXP, '') # remove punctuation
               .tr(' ', '-') # replace spaces with dash
               .squeeze('-') # replace multiple dashes with one
+              .gsub(/\A(\d+)\z/, 'anchor-\1') # digits-only hrefs conflict with issue refs
 
             uniq = headers[id] > 0 ? "-#{headers[id]}" : ''
             headers[id] += 1
diff --git a/spec/lib/banzai/filter/table_of_contents_filter_spec.rb b/spec/lib/banzai/filter/table_of_contents_filter_spec.rb
index 85eddde732e..0cfef4ff5bf 100644
--- a/spec/lib/banzai/filter/table_of_contents_filter_spec.rb
+++ b/spec/lib/banzai/filter/table_of_contents_filter_spec.rb
@@ -65,6 +65,13 @@ describe Banzai::Filter::TableOfContentsFilter do
         expect(doc.css('h2 a').first.attr('href')).to eq '#one-1'
       end
 
+      it 'prepends a prefix to digits-only ids' do
+        doc = filter(header(1, "123") + header(2, "1.0"))
+
+        expect(doc.css('h1 a').first.attr('href')).to eq '#anchor-123'
+        expect(doc.css('h2 a').first.attr('href')).to eq '#anchor-10'
+      end
+
       it 'supports Unicode' do
         doc = filter(header(1, '한글'))
         expect(doc.css('h1 a').first.attr('id')).to eq 'user-content-한글'
diff --git a/spec/lib/gitlab/reference_extractor_spec.rb b/spec/lib/gitlab/reference_extractor_spec.rb
index 476a3f1998d..bce0fa2ea8f 100644
--- a/spec/lib/gitlab/reference_extractor_spec.rb
+++ b/spec/lib/gitlab/reference_extractor_spec.rb
@@ -115,6 +115,15 @@ describe Gitlab::ReferenceExtractor do
     end
   end
 
+  it 'does not include anchors from table of contents in issue references' do
+    issue1 = create(:issue, project: project)
+    issue2 = create(:issue, project: project)
+
+    subject.analyze("not real issue <h4>#{issue1.iid}</h4>, real issue #{issue2.to_reference}")
+
+    expect(subject.issues).to match_array([issue2])
+  end
+
   it 'accesses valid issue objects' do
     @i0 = create(:issue, project: project)
     @i1 = create(:issue, project: project)