various fixes for git and mercurial with InMemoryCommit backend and non-ascii files

- hg use tolocal for unicode objects
- git use core.quotepath=false for commands

rhodecode/controllers/files.py

@@ -306,10 +306,10 b' class FilesController(BaseRepoController'
 
             try:
                 self.scm_model.create_node(repo=c.rhodecode_repo,
-                                             repo_name=repo_name, cs=c.cs,
-                                             user=self.rhodecode_user,
-                                             author=author, message=message,
-                                             content=content, f_path=node_path)
+                                           repo_name=repo_name, cs=c.cs,
+                                           user=self.rhodecode_user,
+                                           author=author, message=message,
+                                           content=content, f_path=node_path)
                 h.flash(_('Successfully committed to %s' % node_path),
                         category='success')
             except NodeAlreadyExistsError, e:

rhodecode/lib/vcs/backends/git/changeset.py

@@ -73,7 +73,6 b' class GitChangeset(BaseChangeset):'
         if ref:
             return safe_unicode(ref)
 
-
     def _fix_path(self, path):
         """
         Paths are stored without trailing slash so we need to get rid off it if
@@ -131,7 +130,6 b' class GitChangeset(BaseChangeset):'
                         name = item
                     self._paths[name] = id
                     self._stat_modes[name] = stat
-
             if not path in self._paths:
                 raise NodeDoesNotExistError("There is no file nor directory "
                     "at the given path %r at revision %r"
@@ -393,7 +391,7 b' class GitChangeset(BaseChangeset):'
     def _diff_name_status(self):
         output = []
         for parent in self.parents:
-            cmd = 'diff --name-status %s %s' % (parent.raw_id, self.raw_id)
+            cmd = 'diff --name-status %s %s --encoding=utf8' % (parent.raw_id, self.raw_id)
             so, se = self.repository.run_git_command(cmd)
             output.append(so.strip())
         return '\n'.join(output)
@@ -409,13 +407,16 b' class GitChangeset(BaseChangeset):'
         for line in self._diff_name_status.splitlines():
             if not line:
                 continue
+
             if line.startswith(char):
-                splitted = line.split(char,1)
+                splitted = line.split(char, 1)
                 if not len(splitted) == 2:
                     raise VCSError("Couldn't parse diff result:\n%s\n\n and "
                         "particularly that line: %s" % (self._diff_name_status,
                         line))
-                paths.add(splitted[1].strip())
+                _path = splitted[1].strip()
+                paths.add(_path)
+
         return sorted(paths)
 
     @LazyProperty

rhodecode/lib/vcs/backends/git/inmemory.py

@@ -5,12 +5,13 b' from dulwich import objects'
 from dulwich.repo import Repo
 from rhodecode.lib.vcs.backends.base import BaseInMemoryChangeset
 from rhodecode.lib.vcs.exceptions import RepositoryError
+from rhodecode.lib.vcs.utils import safe_str
 
 
 class GitInMemoryChangeset(BaseInMemoryChangeset):
 
     def commit(self, message, author, parents=None, branch=None, date=None,
-            **kwargs):
+               **kwargs):
         """
         Performs in-memory commit (doesn't check workdir in any way) and
         returns newly created ``Changeset``. Updates repository's
@@ -120,9 +121,9 b' class GitInMemoryChangeset(BaseInMemoryC'
         commit = objects.Commit()
         commit.tree = commit_tree.id
         commit.parents = [p._commit.id for p in self.parents if p]
-        commit.author = commit.committer = author
+        commit.author = commit.committer = safe_str(author)
         commit.encoding = ENCODING
-        commit.message = message + ' '
+        commit.message = safe_str(message) + ' '
 
         # Compute date
         if date is None:

rhodecode/lib/vcs/backends/git/repository.py

@@ -79,11 +79,13 b' class GitRepository(BaseRepository):'
         :param cmd: git command to be executed
         """
 
-        #cmd = '(cd %s && git %s)' % (self.path, cmd)
+        _copts = ['-c', 'core.quotepath=false', ]
+
         if isinstance(cmd, basestring):
-            cmd = 'GIT_CONFIG_NOGLOBAL=1 git %s' % cmd
-        else:
-            cmd = ['GIT_CONFIG_NOGLOBAL=1', 'git'] + cmd
+            cmd = [cmd]
+
+        cmd = ['GIT_CONFIG_NOGLOBAL=1', 'git'] + _copts + cmd
+
         try:
             opts = dict(
                 shell=isinstance(cmd, basestring),

rhodecode/lib/vcs/backends/hg/inmemory.py

@@ -4,7 +4,7 b' import errno'
 from rhodecode.lib.vcs.backends.base import BaseInMemoryChangeset
 from rhodecode.lib.vcs.exceptions import RepositoryError
 
-from ...utils.hgcompat import memfilectx, memctx, hex
+from ...utils.hgcompat import memfilectx, memctx, hex, tolocal
 
 
 class MercurialInMemoryChangeset(BaseInMemoryChangeset):
@@ -30,9 +30,9 b' class MercurialInMemoryChangeset(BaseInM'
         self.check_integrity(parents)
 
         from .repository import MercurialRepository
-        if not isinstance(message, str) or not isinstance(author, str):
+        if not isinstance(message, unicode) or not isinstance(author, unicode):
             raise RepositoryError('Given message and author needs to be '
-                                  'an <str> instance')
+                                  'an <unicode> instance')
 
         if branch is None:
             branch = MercurialRepository.DEFAULT_BRANCH_NAME
@@ -70,7 +70,7 b' class MercurialInMemoryChangeset(BaseInM'
                         copied=False)
 
             raise RepositoryError("Given path haven't been marked as added,"
-                "changed or removed (%s)" % path)
+                                  "changed or removed (%s)" % path)
 
         parents = [None, None]
         for i, parent in enumerate(self.parents):
@@ -89,9 +89,11 b' class MercurialInMemoryChangeset(BaseInM'
             date=date,
             extra=kwargs)
 
+        loc = lambda u: tolocal(u.encode('utf-8'))
+
         # injecting given _repo params
-        commit_ctx._text = message
-        commit_ctx._user = author
+        commit_ctx._text = loc(message)
+        commit_ctx._user = loc(author)
         commit_ctx._date = date
 
         # TODO: Catch exceptions!

rhodecode/lib/vcs/nodes.py

@@ -120,6 +120,10 b' class Node(object):'
         return None
 
     @LazyProperty
+    def unicode_path(self):
+        return safe_unicode(self.path)
+
+    @LazyProperty
     def name(self):
         """
         Returns name of the node so if its path

rhodecode/lib/vcs/utils/hgcompat.py

@@ -1,6 +1,7 b''
-"""Mercurial libs compatibility
+"""
+Mercurial libs compatibility
+"""
 
-"""
 from mercurial import archival, merge as hg_merge, patch, ui
 from mercurial.commands import clone, nullid, pull
 from mercurial.context import memctx, memfilectx
@@ -10,3 +11,4 b' from mercurial.localrepo import localrep'
 from mercurial.match import match
 from mercurial.mdiff import diffopts
 from mercurial.node import hex
+from mercurial.encoding import tolocal
  No newline at end of file

rhodecode/model/scm.py

@@ -35,7 +35,7 b' from rhodecode.lib.vcs.nodes import File'
 
 from rhodecode import BACKENDS
 from rhodecode.lib import helpers as h
-from rhodecode.lib.utils2 import safe_str
+from rhodecode.lib.utils2 import safe_str, safe_unicode
 from rhodecode.lib.auth import HasRepoPermissionAny, HasReposGroupPermissionAny
 from rhodecode.lib.utils import get_repos as get_filesystem_repos, make_ui, \
     action_logger, EmptyChangeset, REMOVED_REPO_PAT
@@ -369,14 +369,16 b' class ScmModel(BaseModel):'
         # decoding here will force that we have proper encoded values
         # in any other case this will throw exceptions and deny commit
         content = safe_str(content)
-        message = safe_str(message)
         path = safe_str(f_path)
-        author = safe_str(author)
+        # message and author needs to be unicode
+        # proper backend should then translate that into required type
+        message = safe_unicode(message)
+        author = safe_unicode(author)
         m = IMC(repo)
         m.change(FileNode(path, content))
         tip = m.commit(message=message,
-                 author=author,
-                 parents=[cs], branch=cs.branch)
+                       author=author,
+                       parents=[cs], branch=cs.branch)
 
         new_cs = tip.short_id
         action = 'push_local:%s' % new_cs
@@ -403,21 +405,21 b' class ScmModel(BaseModel):'
                 type(content)
             ))
 
-        message = safe_str(message)
+        message = safe_unicode(message)
+        author = safe_unicode(author)
         path = safe_str(f_path)
-        author = safe_str(author)
         m = IMC(repo)
 
         if isinstance(cs, EmptyChangeset):
-            # Emptychangeset means we we're editing empty repository
+            # EmptyChangeset means we we're editing empty repository
             parents = None
         else:
             parents = [cs]
 
         m.add(FileNode(path, content=content))
         tip = m.commit(message=message,
-                 author=author,
-                 parents=parents, branch=cs.branch)
+                       author=author,
+                       parents=parents, branch=cs.branch)
         new_cs = tip.short_id
         action = 'push_local:%s' % new_cs
 

rhodecode/templates/files/files_edit.html

@@ -56,7 +56,7 b''
                       % endif
                     </div>
                 </div>
-                <div class="commit">${_('Editing file')}: ${c.file.path}</div>
+                <div class="commit">${_('Editing file')}: ${c.file.unicode_path}</div>
             </div>
 			    <pre id="editor_pre"></pre>
 				<textarea id="editor" name="content" style="display:none">${h.escape(c.file.content)|n}</textarea>