rst2man does a poor job of creating man pages that can be read on Solaris.
Some of the changes in this patch make the Solaris *roff tools do the right
thing:

  - Getting rid of the rst2man header before the SYNOPSIS

  - Changing "\(aq" to "'", and making sure that single quotes don't start
    a line (at least in a couple of places where it's not safe)

  - Changing ".ft C" and ".ft P" to simply ".ft"

  - Removing "\%" and "\:"

Groff also seems to have as much trouble with the .INDENT/.UNINDENT macros
as Solaris *roff, de-indenting paragraphs that shouldn't be, so we use
.RS/.RE instead.  It's not as fancy, but seems to do the job.

Applying this patch has no effect in the normal userland build (it can't
run automatically until docutils is delivered), but it enables a "gmake
patches/manpage.patch" to regenerate the manpages with the updated text
(found in rst.patch) with the patched hgmanpage.py, and updates the
manpage.patch file.

--- mercurial.hg/doc/hgmanpage.py	2016-10-09 20:00:17.207404849 -0700
+++ mercurial.hg/doc/hgmanpage.py	2016-10-10 10:04:03.695820057 -0700
@@ -202,6 +202,7 @@
         self._in_docinfo = None
         self._active_table = None
         self._in_literal = False
+        self._in_literal_inline = False
         self.header_written = 0
         self._line_block = 0
         self.authors = []
@@ -217,15 +218,15 @@
         # ``B`` bold, ``I`` italic, ``R`` roman should be available.
         # Hopefully ``C`` courier too.
         self.defs = {
-                'indent' : ('.INDENT %.1f\n', '.UNINDENT\n'),
+                'indent' : ('.RS %d\n', '.RE\n'),
                 'definition_list_item' : ('.TP', ''),
                 'field_name' : ('.TP\n.B ', '\n'),
                 'literal' : ('\\fB', '\\fP'),
-                'literal_block' : ('.sp\n.nf\n.ft C\n', '\n.ft P\n.fi\n'),
+                'literal_block' : ('.sp\n.nf\n.ft\n', '\n.ft\n.fi\n'),
 
                 'option_list_item' : ('.TP\n', ''),
 
-                'reference' : (r'\%', r'\:'),
+                'reference' : (r'', r''),
                 'emphasis': ('\\fI', '\\fP'),
                 'strong' : ('\\fB', '\\fP'),
                 'term' : ('\n.B ', '\n'),
@@ -272,6 +273,7 @@
                     self.body[i - 2][:4] == '.TP\n'):
                     self.body[i] = '.\n'
                 elif (self.body[i - 1] == '\n' and
+                    self.body[i - 2] and
                     self.body[i - 2][0] != '.' and
                     (self.body[i - 3][:7] == '.TP\n.B '
                         or self.body[i - 3][:4] == '\n.B ')
@@ -289,7 +291,6 @@
         text = text.replace('\\','\\e')
         replace_pairs = [
             (u'-', u'\\-'),
-            (u"'", u'\\(aq'),
             (u'´', u"\\'"),
             (u'`', u'\\(ga'),
             ]
@@ -302,6 +303,15 @@
             if text[0] == '.':
                 text = '\\&' + text
             text = text.replace('\n.', '\n\\&.')
+        # We need to do the same thing for the inline literals, too
+        if self._in_literal_inline and text[0] == '.':
+            if self.body[-2].endswith('\n') and \
+              self.body[-1] == self.defs['literal'][0]:
+                  self.body.append('\\&')
+        # Single quotes starting a line are harmful, too.
+        if text[0] == "'":
+                text = '\\&' + text
+        text = text.replace("\n'", "\n\\&'")
         self.body.append(text)
 
     def depart_Text(self, node):
@@ -386,7 +396,6 @@
         if self.header_written:
             return
         self.body.append(self.header())
-        self.body.append(MACRO_DEF)
         self.header_written = 1
 
     def visit_address(self, node):
@@ -800,8 +809,10 @@
 
     def visit_literal(self, node):
         self.body.append(self.defs['literal'][0])
+        self._in_literal_inline = True
 
     def depart_literal(self, node):
+        self._in_literal_inline = False
         self.body.append(self.defs['literal'][1])
 
     def visit_literal_block(self, node):