Changeset - 7268209e8843
[Not reviewed]
default
0 1 0
Mads Kiilerich - 7 years ago 2019-01-07 00:00:22
mads@kiilerich.com
Grafted from: e33e6926d29e
middleware: drop special handling of hg RepoError

The 'not found' case should never happen - we check that repo exists before
calling out to hg.

And the Mercurial RepoError handling was muting if *not* 'not found'. That
seems very harmful - better to get a clear error.

We are thus better without the special handling of RepoError.
1 file changed with 0 insertions and 3 deletions:
0 comments (0 inline, 0 general)
kallithea/lib/middleware/simplehg.py
Show inline comments
 
@@ -58,180 +58,177 @@ def is_mercurial(environ):
 
        ishg_path = False
 

	
 
    log.debug('pathinfo: %s detected as Mercurial %s',
 
        path_info, ishg_path
 
    )
 
    return ishg_path
 

	
 

	
 
def get_header_hgarg(environ):
 
    """Decode the special Mercurial encoding of big requests over multiple headers.
 
    >>> get_header_hgarg({})
 
    ''
 
    >>> get_header_hgarg({'HTTP_X_HGARG_0': ' ', 'HTTP_X_HGARG_1': 'a','HTTP_X_HGARG_2': '','HTTP_X_HGARG_3': 'b+c %20'})
 
    'ab+c %20'
 
    """
 
    chunks = []
 
    i = 1
 
    while True:
 
        v = environ.get('HTTP_X_HGARG_%d' % i)
 
        if v is None:
 
            break
 
        chunks.append(v)
 
        i += 1
 
    return ''.join(chunks)
 

	
 

	
 
class SimpleHg(BaseVCSController):
 

	
 
    def _handle_request(self, environ, start_response):
 
        if not is_mercurial(environ):
 
            return self.application(environ, start_response)
 

	
 
        ip_addr = self._get_ip_addr(environ)
 
        # skip passing error to error controller
 
        environ['pylons.status_code_redirect'] = True
 

	
 
        #======================================================================
 
        # EXTRACT REPOSITORY NAME FROM ENV
 
        #======================================================================
 
        try:
 
            str_repo_name = self.__get_repository(environ)
 
            repo_name = safe_unicode(str_repo_name)
 
            log.debug('Extracted repo name is %s', repo_name)
 
        except Exception as e:
 
            log.error('error extracting repo_name: %r', e)
 
            return HTTPInternalServerError()(environ, start_response)
 

	
 
        # quick check if that dir exists...
 
        if not is_valid_repo(repo_name, self.basepath, 'hg'):
 
            return HTTPNotFound()(environ, start_response)
 

	
 
        #======================================================================
 
        # GET ACTION PULL or PUSH
 
        #======================================================================
 
        try:
 
            action = self.__get_action(environ)
 
        except HTTPBadRequest as e:
 
            return e(environ, start_response)
 

	
 
        #======================================================================
 
        # CHECK PERMISSIONS
 
        #======================================================================
 
        user, response_app = self._authorize(environ, start_response, action, repo_name, ip_addr)
 
        if response_app is not None:
 
            return response_app(environ, start_response)
 

	
 
        # extras are injected into Mercurial UI object and later available
 
        # in hooks executed by Kallithea
 
        from kallithea import CONFIG
 
        server_url = get_server_url(environ)
 
        extras = {
 
            'ip': ip_addr,
 
            'username': user.username,
 
            'action': action,
 
            'repository': repo_name,
 
            'scm': 'hg',
 
            'config': CONFIG['__file__'],
 
            'server_url': server_url,
 
        }
 
        #======================================================================
 
        # MERCURIAL REQUEST HANDLING
 
        #======================================================================
 
        repo_path = os.path.join(safe_str(self.basepath), str_repo_name)
 
        log.debug('Repository path is %s', repo_path)
 

	
 
        fix_PATH()
 
        log.debug('HOOKS extras is %s', extras)
 
        baseui = make_ui(repo_path=repo_path)
 
        _set_extras(extras or {})
 

	
 
        try:
 
            log.info('%s action on Mercurial repo "%s" by "%s" from %s',
 
                     action, str_repo_name, safe_str(user.username), ip_addr)
 
            environ['REPO_NAME'] = str_repo_name # used by hgweb_mod.hgweb
 
            app = self.__make_app(repo_path, baseui)
 
            return app(environ, start_response)
 
        except RepoError as e:
 
            if str(e).find('not found') != -1:
 
                return HTTPNotFound()(environ, start_response)
 
        except Exception:
 
            log.error(traceback.format_exc())
 
            return HTTPInternalServerError()(environ, start_response)
 

	
 
    def __make_app(self, repo_name, baseui):
 
        """
 
        Make an hgweb wsgi application using baseui.
 
        """
 
        return hgweb_mod.hgweb(repo_name, name=repo_name, baseui=baseui)
 

	
 
    def __get_repository(self, environ):
 
        """
 
        Gets repository name out of PATH_INFO header
 

	
 
        :param environ: environ where PATH_INFO is stored
 
        """
 
        try:
 
            environ['PATH_INFO'] = self._get_by_id(environ['PATH_INFO'])
 
            repo_name = '/'.join(environ['PATH_INFO'].split('/')[1:])
 
            if repo_name.endswith('/'):
 
                repo_name = repo_name.rstrip('/')
 
        except Exception:
 
            log.error(traceback.format_exc())
 
            raise
 

	
 
        return repo_name
 

	
 
    def __get_action(self, environ):
 
        """
 
        Maps Mercurial request commands into 'pull' or 'push'.
 

	
 
        Raises HTTPBadRequest if the request environment doesn't look like a hg client.
 
        """
 
        mapping = {
 
            # 'batch' is not in this list - it is handled explicitly
 
            'between': 'pull',
 
            'branches': 'pull',
 
            'branchmap': 'pull',
 
            'capabilities': 'pull',
 
            'changegroup': 'pull',
 
            'changegroupsubset': 'pull',
 
            'changesetdata': 'pull',
 
            'clonebundles': 'pull',
 
            'debugwireargs': 'pull',
 
            'filedata': 'pull',
 
            'getbundle': 'pull',
 
            'getlfile': 'pull',
 
            'heads': 'pull',
 
            'hello': 'pull',
 
            'known': 'pull',
 
            'lheads': 'pull',
 
            'listkeys': 'pull',
 
            'lookup': 'pull',
 
            'manifestdata': 'pull',
 
            'narrow_widen': 'pull',
 
            'protocaps': 'pull',
 
            'statlfile': 'pull',
 
            'stream_out': 'pull',
 
            'pushkey': 'push',
 
            'putlfile': 'push',
 
            'unbundle': 'push',
 
            }
 
        for qry in environ['QUERY_STRING'].split('&'):
 
            parts = qry.split('=', 1)
 
            if len(parts) == 2 and parts[0] == 'cmd':
 
                cmd = parts[1]
 
                if cmd == 'batch':
 
                    hgarg = get_header_hgarg(environ)
 
                    if not hgarg.startswith('cmds='):
 
                        return 'push' # paranoid and safe
 
                    for cmd_arg in hgarg[5:].split(';'):
 
                        cmd, _args = urllib.unquote_plus(cmd_arg).split(' ', 1)
 
                        op = mapping.get(cmd, 'push')
 
                        if op != 'pull':
 
                            assert op == 'push'
 
                            return 'push'
 
                    return 'pull'
 
                return mapping.get(cmd, 'push')
 

	
 
        # Note: the client doesn't get the helpful error message
 
        raise HTTPBadRequest('Unable to detect pull/push action! Are you using non standard command or client?')
0 comments (0 inline, 0 general)