From patchwork Thu Nov  2 13:32:53 2017
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Subject: patch: improve heuristics to not take the word "diff" as header
 (issue1879)
From: Yuya Nishihara <yuya@tcha.org>
X-Patchwork-Id: 25344
Message-Id: <6798568fe43dc1e28a0c.1509629573@mimosa>
To: mercurial-devel@mercurial-scm.org
Date: Thu, 02 Nov 2017 22:32:53 +0900

# HG changeset patch
# User Yuya Nishihara <yuya@tcha.org>
# Date 1508572257 -32400
#      Sat Oct 21 16:50:57 2017 +0900
# Node ID 6798568fe43dc1e28a0c66c54217b563e1f2d5de
# Parent  a7e49a5b3e6f38338660c0d98f9c593f70064e95
patch: improve heuristics to not take the word "diff" as header (issue1879)

The word "diff" is likely to appear in a commit message. Let's make it less
likely by requiring "diff -" for "diff -r" or "diff --git".

diff --git a/mercurial/patch.py b/mercurial/patch.py
--- a/mercurial/patch.py
+++ b/mercurial/patch.py
@@ -203,7 +203,7 @@ def extract(ui, fileobj):
 
     # attempt to detect the start of a patch
     # (this heuristic is borrowed from quilt)
-    diffre = re.compile(br'^(?:Index:[ \t]|diff[ \t]|RCS file: |'
+    diffre = re.compile(br'^(?:Index:[ \t]|diff[ \t]-|RCS file: |'
                         br'retrieving revision [0-9]+(\.[0-9]+)*$|'
                         br'---[ \t].*?^\+\+\+[ \t]|'
                         br'\*\*\*[ \t].*?^---[ \t])',
diff --git a/tests/test-import.t b/tests/test-import.t
--- a/tests/test-import.t
+++ b/tests/test-import.t
@@ -1346,6 +1346,93 @@ import a unified diff with no lines of c
 
   $ cd ..
 
+commit message that looks like a diff header (issue1879)
+
+  $ hg init headerlikemsg
+  $ cd headerlikemsg
+  $ touch empty
+  $ echo nonempty >> nonempty
+  $ hg ci -qAl - <<EOF
+  > blah blah
+  > diff blah
+  > blah blah
+  > EOF
+  $ hg --config diff.git=1 log -pv
+  changeset:   0:c6ef204ef767
+  tag:         tip
+  user:        test
+  date:        Thu Jan 01 00:00:00 1970 +0000
+  files:       empty nonempty
+  description:
+  blah blah
+  diff blah
+  blah blah
+  
+  
+  diff --git a/empty b/empty
+  new file mode 100644
+  diff --git a/nonempty b/nonempty
+  new file mode 100644
+  --- /dev/null
+  +++ b/nonempty
+  @@ -0,0 +1,1 @@
+  +nonempty
+  
+
+ (without --git, empty file is lost, but commit message should be preserved)
+
+  $ hg init plain
+  $ hg export 0 | hg -R plain import -
+  applying patch from stdin
+  $ hg --config diff.git=1 -R plain log -pv
+  changeset:   0:60a2d231e71f
+  tag:         tip
+  user:        test
+  date:        Thu Jan 01 00:00:00 1970 +0000
+  files:       nonempty
+  description:
+  blah blah
+  diff blah
+  blah blah
+  
+  
+  diff --git a/nonempty b/nonempty
+  new file mode 100644
+  --- /dev/null
+  +++ b/nonempty
+  @@ -0,0 +1,1 @@
+  +nonempty
+  
+
+ (with --git, patch contents should be fully preserved)
+
+  $ hg init git
+  $ hg --config diff.git=1 export 0 | hg -R git import -
+  applying patch from stdin
+  $ hg --config diff.git=1 -R git log -pv
+  changeset:   0:c6ef204ef767
+  tag:         tip
+  user:        test
+  date:        Thu Jan 01 00:00:00 1970 +0000
+  files:       empty nonempty
+  description:
+  blah blah
+  diff blah
+  blah blah
+  
+  
+  diff --git a/empty b/empty
+  new file mode 100644
+  diff --git a/nonempty b/nonempty
+  new file mode 100644
+  --- /dev/null
+  +++ b/nonempty
+  @@ -0,0 +1,1 @@
+  +nonempty
+  
+
+  $ cd ..
+
 no segfault while importing a unified diff which start line is zero but chunk
 size is non-zero