Fix amending commits with Unicode command line arguments

Author: sds

CHANGELOG.md

@@ -1,5 +1,10 @@
 # Overcommit Changelog
 
+## master (unreleased)
+
+* Fix bug where amending a commit with command line arguments containing
+  Unicode characters could cause a crash due to invalid byte sequences
+
 ## 0.32.0.rc1
 
 * Add `concurrency` global option allowing you to specify the number of threads

lib/overcommit/hook_context/pre_commit.rb

@@ -15,6 +15,13 @@ def amendment?
       cmd = Overcommit::Utils.parent_command
       amend_pattern = 'commit(\s.*)?\s--amend(\s|$)'
 
+      # Since the ps command can return invalid byte sequences for commands
+      # containing unicode characters, we replace the offending characters,
+      # since the pattern we're looking for will consist of ASCII characters
+      unless cmd.valid_encoding?
+        cmd = cmd.encode('UTF-16be', invalid: :replace, replace: '?').encode('UTF-8')
+      end
+
       return @amendment if
         # True if the command is a commit with the --amend flag
         @amendment = !(/\s#{amend_pattern}/ =~ cmd).nil?

spec/overcommit/hook_context/pre_commit_spec.rb

@@ -20,6 +20,12 @@
       it { should == true }
     end
 
+    context 'when the parent command contains invalid byte sequence' do
+      let(:command) { "git commit --amend -m \xE3M^AM^B" }
+
+      it { should == true }
+    end
+
     context 'when amending a commit using a git alias' do
       around do |example|
         repo do