[PATCH 2 of 4 py3] revsetlang: work around repr() returning unicode on py3

[Augie Fackler]

# HG changeset patch
# User Augie Fackler <augie at google.com>
# Date 1489900459 14400
#      Sun Mar 19 01:14:19 2017 -0400
# Node ID 63466a54ec9e266031cce12fbd699ccfccd40738
# Parent  a4745fd9219ed5b408bfc0403a4a8e6acd41df6c
revsetlang: work around repr() returning unicode on py3

I'm not confident in my choice of "encode to utf8" here: it seems
reasonable that we could expect the revset engine to do the right
thing there. On the other hand these strings might have encoding
weirdness to begin with if there's `branch(some-weird-byte-sequence)`.

diff --git a/mercurial/revsetlang.py b/mercurial/revsetlang.py
--- a/mercurial/revsetlang.py
+++ b/mercurial/revsetlang.py
@@ -607,7 +607,11 @@ def formatspec(expr, *args):
     '''
 
     def quote(s):
-        return repr(str(s))
+        r = repr(bytes(s))
+        if pycompat.ispy3:
+            r = r[1:] # strip off b prefix
+            return r.encode('utf-8')
+        return r
 
     def argtype(c, arg):
         if c == 'd':