logilab-astng #92362 Pyreverse crashes while scanning anaconda project [validation pending]

Comments

• 2012/04/24 13:05, written by acampeas

  can you give the full line and your current working directory ?
• 2012/04/24 13:13, written by vorcak.jan@gmail.com

  I've added log file http://www.logilab.org/file/92364
  
  • 2012/04/24 13:20, written by acampeas

    ok, then what if you invoque pyreverse from an upper directory ?

    eg:

    cd /home/xvorcak/anaconda
    pyreverse pyanaconda
    

    ?
    • 2012/04/24 13:25, written by vorcak.jan@gmail.com

      the same thing happened http://www.logilab.org/file/92368
      
      • 2012/04/24 13:34, written by acampeas

        Interestingly it's not the same thing (crash happens in a different place).

        Can you try a last time with the following patch applied to logilab.common ?

        diff --git a/modutils.py b/modutils.py
        --- a/modutils.py
        +++ b/modutils.py
        @@ -331,7 +331,7 @@
             starti = 0
             if parts[0] == '':
                 assert context_file is not None, \
        -                'explicit relative import, but no context_file?'
        +                'explicit relative import (%s), but no context_file?' % dotted_name
                 path = [] # prevent resolving the import non-relatively
                 starti = 1
             while parts[starti] == '': # for all further dots: change context
        
        • 2012/04/24 13:52, written by vorcak.jan@gmail.com

          it fails on module .DeviceFormat ... 'explicit relative import (%s), but no context_file?' % dotted_name AssertionError: explicit relative import (.DeviceFormat), but no context_file? // This module is located in pyanaconda/storage/formats/__init__.py
          • 2012/04/24 14:03, written by acampeas

            So there would be a " .DeviceFormat " module ? or package ?

            Can you have a look at the exact nature of this ?
            • 2012/04/24 14:12, written by vorcak.jan@gmail.com

              I think it's solved by sthenault's patch because context_file has not been specified
              
              • 2012/04/24 14:19, written by acampeas

                Yes, I've seen it. This is likely the "right fix". I've now lost all hope to understand what's going on ...

                Unless you can answer ma question, maybe ... ;-)
• 2012/04/24 13:59, written by sthenault

  the pb is simply that Linker.visit_from call get_module_part without specifying a context_file.

  Try:

  diff --git a/inspector.py b/inspector.py
  --- a/inspector.py
  +++ b/inspector.py
  @@ -252,11 +252,11 @@ class Linker(IdGeneratorMixIn, LocalsVis
              # analyze dependencies
              fullname = '%s.%s' % (basename, name[0])
              if fullname.find('.') > -1:
                  try:
                      # XXX: don't use get_module_part, missing package precedence
  -                    fullname = get_module_part(fullname)
  +                    fullname = get_module_part(fullname, node.root().file)
                  except ImportError:
                      continue
              if fullname != basename:
                  self._imported_module(node, fullname, relative)
  
  • 2012/04/24 14:11, written by vorcak.jan@gmail.com

    this one works, thank you, can I ask when I can expect this to be fixed in a repositories?
    
    • 2012/04/24 14:47, written by sthenault

      soon :)
      • 2012/08/16 07:38, written by malte.vesper

        Please add it, would have saved you one unnecessary bug report by me ;). Or are there issues with the patch?
        
        • 2012/08/16 21:13, written by lothiraldan

          It's already in repository (see http://hg.logilab.org/logilab/astng/rev/27ca2820b32b) but I was not included in previous release (for various reasons).
          Will be available in next release.