Introduce a rendering limit for commit titles

We pass "full" commit titles through Banzai in just one place:
Projects::RefsController#logs_tree (via Gitlab::TreeSummary). It is
shown (truncated to 100 characters, with the rest in tooltips) on the
repository file browser.

Permitting unlimited-length lines to be passed through banzai is a
recipe for puma timeouts, so add a 1KiB limit to these, mirroring the
1MiB limit we added to commit descriptions.

We also clean up the omission string used (from unicode to '...') so it
renders correctly in more places.

app/models/commit.rb

@@ -36,7 +36,7 @@ class Commit
   LINK_EXTENSION_PATTERN = /(patch)/.freeze
 
   cache_markdown_field :title, pipeline: :single_line
-  cache_markdown_field :full_title, pipeline: :single_line
+  cache_markdown_field :full_title, pipeline: :single_line, limit: 1.kilobyte
   cache_markdown_field :description, pipeline: :commit_description, limit: 1.megabyte
 
   class << self

changelogs/unreleased/remove-banzai-commit-full-title.yml

+---
+title: Introduce a rendering limit for commit titles
+merge_request: 52904
+author:
+type: performance

doc/administration/instance_limits.md

@@ -127,6 +127,15 @@ It's possible that this limit will be changed to a lower number in the future.
 
 - **Max size:** ~1 million characters / ~1 MB
 
+## Size of commit titles and descriptions
+
+> [Introduced](https://gitlab.com/gitlab-org/gitlab/-/issues/292039) in GitLab 13.9
+
+Commits with arbitrarily large messages may be pushed to GitLab, but when
+displaying commits, titles (the first line of the commit message) will be
+limited to 1KiB, and descriptions (the rest of the message) will be limited to
+1MiB.
+
 ## Number of issues in the milestone overview
 
 > [Introduced](https://gitlab.com/gitlab-org/gitlab/-/issues/39453) in GitLab 12.10.

lib/banzai/filter/truncate_source_filter.rb

@@ -6,7 +6,9 @@ module Banzai
       def call
         return text unless context.key?(:limit)
 
-        text.truncate_bytes(context[:limit])
+        # Use three dots instead of the ellipsis Unicode character because
+        # some clients show the raw Unicode value in the merge commit.
+        text.truncate_bytes(context[:limit], omission: '...')
       end
     end
   end

spec/lib/banzai/filter/truncate_source_filter_spec.rb

@@ -22,7 +22,7 @@ RSpec.describe Banzai::Filter::TruncateSourceFilter do
 
   it 'truncates UTF-8 text by bytes, on a character boundary' do
     utf8_text = '日本語の文字が大きい'
-    truncated = '日…'
+    truncated = '日...'
 
     expect(filter(utf8_text, limit: truncated.bytesize)).to eq(truncated)
     expect(filter(utf8_text, limit: utf8_text.bytesize)).to eq(utf8_text)

spec/lib/banzai/pipeline/pre_process_pipeline_spec.rb

@@ -30,6 +30,6 @@ RSpec.describe Banzai::Pipeline::PreProcessPipeline do
 
     result = described_class.call(text, limit: 12)
 
-    expect(result[:output]).to eq('foo foo f…')
+    expect(result[:output]).to eq('foo foo f...')
   end
 end

spec/models/commit_spec.rb

@@ -400,6 +400,19 @@ eos
       allow(commit).to receive(:safe_message).and_return(message + "\n" + message)
       expect(commit.full_title).to eq(message)
     end
+
+    it 'truncates html representation if more than 1KiB' do
+      # Commit title is over 2KiB on a single line
+      huge_commit_title = ('panic ' * 350) + 'trailing text'
+
+      allow(commit).to receive(:safe_message).and_return(huge_commit_title)
+
+      commit.refresh_markdown_cache
+      full_title_html = commit.full_title_html
+
+      expect(full_title_html.bytesize).to be < 2.kilobytes
+      expect(full_title_html).not_to include('trailing text')
+    end
   end
 
   describe 'description' do