Merge branch '21247-mergerequestscontroller-conflicts-may-fail-with-iso-8859-data' into 'master'

Handle non-UTF-8 conflicts gracefully

## What does this MR do?

If a conflict file isn't in a UTF-8-compatible encoding, we can't resolve it in the UI.

## What are the relevant issue numbers?

Closes #21247. 

See merge request !5961

CHANGELOG

@@ -26,6 +26,9 @@ v 8.11.2
   - Don't show resolve conflicts link before MR status is updated
   - Don't prevent viewing the MR when git refs for conflicts can't be found on disk
 
+v 8.11.3 (unreleased)
+  - Don't try to show merge conflict resolution info if a merge conflict contains non-UTF-8 characters
+
 v 8.11.2 (unreleased)
   - Show "Create Merge Request" widget for push events to fork projects on the source project
 

doc/user/project/merge_requests/resolve_conflicts.md

@@ -26,6 +26,7 @@ this is similar to performing `git checkout feature; git merge master` locally.
 GitLab allows resolving conflicts in a file where all of the below are true:
 
 - The file is text, not binary
+- The file is in a UTF-8 compatible encoding
 - The file does not already contain conflict markers
 - The file, with conflict markers added, is not over 200 KB in size
 - The file exists under the same path in both branches

lib/gitlab/conflict/parser.rb

@@ -13,10 +13,19 @@ module Gitlab
       class UnmergeableFile < ParserError
       end
 
+      class UnsupportedEncoding < ParserError
+      end
+
       def parse(text, our_path:, their_path:, parent_file: nil)
         raise UnmergeableFile if text.blank? # Typically a binary file
         raise UnmergeableFile if text.length > 102400
 
+        begin
+          text.to_json
+        rescue Encoding::UndefinedConversionError
+          raise UnsupportedEncoding
+        end
+
         line_obj_index = 0
         line_old = 1
         line_new = 1

spec/features/merge_requests/conflicts_spec.rb

@@ -43,7 +43,8 @@ feature 'Merge request conflict resolution', js: true, feature: true do
     'conflict-too-large' => 'when the conflicts contain a large file',
     'conflict-binary-file' => 'when the conflicts contain a binary file',
     'conflict-contains-conflict-markers' => 'when the conflicts contain a file with ambiguous conflict markers',
-    'conflict-missing-side' => 'when the conflicts contain a file edited in one branch and deleted in another'
+    'conflict-missing-side' => 'when the conflicts contain a file edited in one branch and deleted in another',
+    'conflict-non-utf8' => 'when the conflicts contain a non-UTF-8 file',
   }
 
   UNRESOLVABLE_CONFLICTS.each do |source_branch, description|

spec/lib/gitlab/conflict/parser_spec.rb

@@ -183,6 +183,11 @@ CONFLICT
         expect { parse_text('a' * 102401) }.
           to raise_error(Gitlab::Conflict::Parser::UnmergeableFile)
       end
+
+      it 'raises UnsupportedEncoding when the file contains non-UTF-8 characters' do
+        expect { parse_text("a\xC4\xFC".force_encoding(Encoding::ASCII_8BIT)) }.
+          to raise_error(Gitlab::Conflict::Parser::UnsupportedEncoding)
+      end
     end
   end
 end

spec/support/test_env.rb

@@ -24,11 +24,12 @@ module TestEnv
     'expand-collapse-lines'              => '238e82d',
     'video'                              => '8879059',
     'crlf-diff'                          => '5938907',
-    'conflict-start'                     => '14fa46b',
+    'conflict-start'                     => '75284c7',
     'conflict-resolvable'                => '1450cd6',
     'conflict-binary-file'               => '259a6fb',
     'conflict-contains-conflict-markers' => '5e0964c',
     'conflict-missing-side'              => 'eb227b3',
+    'conflict-non-utf8'                  => 'd0a293c',
     'conflict-too-large'                 => '39fa04f',
   }