filectx: fix cmp() of file starting with '\1\n' stable

filectx: fix cmp() of file starting with '\1\n'

If file data starts with '\1\n', it will be escaped in the revlog to
create an empty metadata block, thus adding four bytes to the size in
the revlog size index. There's no way to detect that this has happened
in filelog.size() faster than decompressing each revision [1].

For filectx.cmp(), we have the size of the file in the working directory
available. If it differs by exactly four bytes, it may be this case, so
do a full comparison.

 [1]: http://markmail.org/message/5akdbmmqx7vq2fsg

--- a/mercurial/context.py	Wed Jan 11 09:26:47 2012 -0600
+++ b/mercurial/context.py	Thu Jan 12 00:49:45 2012 +0900
@@ -368,7 +368,11 @@
 
         returns True if different than fctx.
         """
-        if (fctx._filerev is None and self._repo._encodefilterpats
+        if (fctx._filerev is None
+            and (self._repo._encodefilterpats
+                 # if file data starts with '\1\n', empty metadata block is
+                 # prepended, which adds 4 bytes to fielog.size().
+                 or self.size() - 4 == fctx.size())
             or self.size() == fctx.size()):
             return self._filelog.cmp(self._filenode, fctx.data())
 

--- a/tests/test-status.t	Wed Jan 11 09:26:47 2012 -0600
+++ b/tests/test-status.t	Thu Jan 12 00:49:45 2012 +0900
@@ -272,3 +272,26 @@
     modified
   R removed
   C deleted
+
+  $ cd ..
+
+hg status of binary file starting with '\1\n', a separator for metadata:
+
+  $ hg init repo5
+  $ cd repo5
+  $ printf '\1\nfoo' > 010a
+  $ hg ci -q -A -m 'initial checkin'
+  $ hg status -A
+  C 010a
+
+  $ printf '\1\nbar' > 010a
+  $ hg status -A
+  M 010a
+  $ hg ci -q -m 'modify 010a'
+  $ hg status -A --rev 0:1
+  M 010a
+
+  $ touch empty
+  $ hg ci -q -A -m 'add another file'
+  $ hg status -A --rev 1:2 010a
+  C 010a