diff hgext/extdiff.py @ 23680:4075f2f8ea53

extdiff: avoid unexpected quoting arguments for external tools (issue4463)

Before this patch, all command line arguments for external tools are
quoted by the combination of "shlex.split" and "util.shellquote". But
this causes some problems.

  - some problematic commands can't work correctly with quoted arguments

    For example, 'WinMerge /r ....' is OK, but 'WinMerge "/r" ....' is
    NG. See also below for detail about this problem.

        https://bitbucket.org/tortoisehg/thg/issue/3978/

  - quoting itself may change semantics of arguments

    For example, when the environment variable CONCAT="foo bar baz':

      - mydiff $CONCAT   => mydiff foo bar baz   (taking 3 arguments)
      - mydiff "$CONCAT" => mydiff "foo bar baz" (taking only 1 argument)

    For another example, single quoting (= "util.shellquote") on POSIX
    environment prevents shells from expanding environment variables,
    tilde, and so on:

      - mydiff "$HOME" => mydiff /home/foobar
      - mydiff '$HOME' => mydiff $HOME

  - "shlex.split" can't handle some special characters correctly

    It just splits specified command line by whitespaces.

    For example, "echo foo;echo bar" is split into ["echo",
    "foo;echo", "bar"].

On the other hand, if quoting itself is omitted, users can't specify
options including space characters with "--option" at runtime.

The root cause of this issue is that "shlex.split + util.shellquote"
combination loses whether users really want to quote each command line
elements or not, even though these can be quoted arbitrarily in
configurations.

To resolve this problem, this patch does:

  - prevent configurations from being processed by "shlex.split" and
    "util.shellquote"

    only (possibly) "findexe"-ed or "findexternaltool"-ed command path
    is "util.shellquote", because it may contain whitespaces.

  - quote options specified by "--option" via command line at runtime

This patch also makes "dodiff()" take only one "args" argument instead
of "diffcmd" and "diffopts. It also omits applying "util.shellquote"
on "args", because "args" should be already stringified in "extdiff()"
and "mydiff()".

The last hunk for "test-extdiff.t" replaces two whitespaces by single
whitespace, because change of "' '.join()" logic causes omitting
redundant whitespaces.

--- a/hgext/extdiff.py	Sun Dec 28 23:59:57 2014 +0100
+++ b/hgext/extdiff.py	Thu Dec 25 23:33:26 2014 +0900
@@ -109,7 +109,7 @@
                                   os.lstat(dest).st_mtime))
     return dirname, fns_and_mtime
 
-def dodiff(ui, repo, diffcmd, diffopts, pats, opts):
+def dodiff(ui, repo, args, pats, opts):
     '''Do the actual diff:
 
     - copy to a temp structure if diffing 2 internal revisions
@@ -120,7 +120,6 @@
 
     revs = opts.get('rev')
     change = opts.get('change')
-    args = ' '.join(map(util.shellquote, diffopts))
     do3way = '$parent2' in args
 
     if revs and change:
@@ -222,8 +221,7 @@
         regex = '\$(parent2|parent1?|child|plabel1|plabel2|clabel|root)'
         if not do3way and not re.search(regex, args):
             args += ' $parent1 $child'
-        args = re.sub(regex, quote, args)
-        cmdline = util.shellquote(diffcmd) + ' ' + args
+        cmdline = re.sub(regex, quote, args)
 
         ui.debug('running %r in %s\n' % (cmdline, tmproot))
         ui.system(cmdline, cwd=tmproot)
@@ -271,7 +269,8 @@
     if not program:
         program = 'diff'
         option = option or ['-Npru']
-    return dodiff(ui, repo, program, option, pats, opts)
+    cmdline = ' '.join(map(util.shellquote, [program] + option))
+    return dodiff(ui, repo, cmdline, pats, opts)
 
 def uisetup(ui):
     for cmd, path in ui.configitems('extdiff'):
@@ -281,29 +280,37 @@
                 path = util.findexe(cmd)
                 if path is None:
                     path = filemerge.findexternaltool(ui, cmd) or cmd
-            diffopts = shlex.split(ui.config('extdiff', 'opts.' + cmd, ''))
+            diffopts = ui.config('extdiff', 'opts.' + cmd, '')
+            cmdline = util.shellquote(path)
+            if diffopts:
+                cmdline += ' ' + diffopts
         elif cmd.startswith('opts.'):
             continue
         else:
-            # command = path opts
             if path:
-                diffopts = shlex.split(path)
-                path = diffopts.pop(0)
+                # case "cmd = path opts"
+                cmdline = path
+                diffopts = len(shlex.split(cmdline)) > 1
             else:
-                path, diffopts = util.findexe(cmd), []
+                # case "cmd ="
+                path = util.findexe(cmd)
                 if path is None:
                     path = filemerge.findexternaltool(ui, cmd) or cmd
+                cmdline = util.shellquote(path)
+                diffopts = False
         # look for diff arguments in [diff-tools] then [merge-tools]
-        if diffopts == []:
+        if not diffopts:
             args = ui.config('diff-tools', cmd+'.diffargs') or \
                    ui.config('merge-tools', cmd+'.diffargs')
             if args:
-                diffopts = shlex.split(args)
-        def save(cmd, path, diffopts):
+                cmdline += ' ' + args
+        def save(cmdline):
             '''use closure to save diff command to use'''
             def mydiff(ui, repo, *pats, **opts):
-                return dodiff(ui, repo, path, diffopts + opts['option'],
-                              pats, opts)
+                options = ' '.join(map(util.shellquote, opts['option']))
+                if options:
+                    options = ' ' + options
+                return dodiff(ui, repo, cmdline + options, pats, opts)
             doc = _('''\
 use %(path)s to diff repository (or selected files)
 
@@ -325,6 +332,6 @@
             # right encoding) prevents that.
             mydiff.__doc__ = doc.decode(encoding.encoding)
             return mydiff
-        cmdtable[cmd] = (save(cmd, path, diffopts),
+        cmdtable[cmd] = (save(cmdline),
                          cmdtable['extdiff'][1][1:],
                          _('hg %s [OPTION]... [FILE]...') % cmd)