Traceback merging python packages

[masterdriverz]

I get this traceback merging some python packages (docutils and elementtree so far)

>>> Replacing dev-python/elementtree-1.2.6 with dev-python/elementtree-1.2.6-r2
Traceback (most recent call last):
  File "/usr/bin/pmerge", line 36, in <module>
    commandline.main(subcommands)
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/util/commandline.py", line 401, in main
    exitstatus = main_func(options, out, err)
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/scripts/pmerge.py", line 548, in main
    ret = i.finish()
  File "/home/chaza/bzr/snakeoil/integration/snakeoil/dependant_methods.py", line 26, in ensure_deps
    r = getattr(self, dep).raw_func(*a, **kw)
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/interfaces/repo.py", line 309, in start
    return livefs_base.start(self, engine)
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/interfaces/repo.py", line 118, in start
    self.me.sanity_check()
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/merge/engine.py", line 277, in execute_hook
    trigger(self, self.csets)
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/merge/triggers.py", line 120, in __call__
    return self.trigger(engine, *self._get_csets(required_csets, csets))
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/merge/triggers.py", line 108, in _get_csets
    return [csets[x] for x in required_csets]
  File "/home/chaza/bzr/snakeoil/integration/snakeoil/mappings.py", line 345, in __getitem__
    return x[key]
  File "/home/chaza/bzr/snakeoil/integration/snakeoil/mappings.py", line 173, in __getitem__
    v = self._vals[key] = self._val_func(key)
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/merge/engine.py", line 217, in _get_cset_source
    return self.cset_sources[key](self, self.csets)
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/merge/engine.py", line 310, in get_install_livefs_intersect
    return engine._get_livefs_intersect_cset(engine, csets, "install")
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/merge/engine.py", line 306, in _get_livefs_intersect_cset
    return contents.contentsSet(livefs.intersect(csets[cset_name]))
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/fs/contents.py", line 40, in __init__
    self._dict.update(check_instance(x) for x in initial)
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/fs/contents.py", line 40, in <genexpr>
    self._dict.update(check_instance(x) for x in initial)
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/fs/livefs.py", line 149, in intersect
    yield f(x.location)
  File "/home/chaza/bzr/pkgcore/integration/pkgcore/fs/livefs.py", line 45, in gen_obj
    stat = os.lstat(real_location)
OSError: [Errno 20] Not a directory: '/usr/lib/python2.5/site-packages/elementtree-1.2.6_20050316-py2.5.egg-info/top_level.txt'

[jokey]

this is supposed to be gone long ago

[masterdriverz]

This is still alive and kicking, not sure why, but it is. For some reason random files are being appended to .egg-info paths, and the .egg-info paths are being mistaken for directories.

[masterdriverz]

Replying to masterdriverz:

This is still alive and kicking, not sure why, but it is. For some reason random files are being appended to .egg-info paths, and the .egg-info paths are being mistaken for directories.

Well that was completely wrong. Basically /usr/lib/python2.5/site-packages/docutils-0.4-py2.5.egg-info is a file in one version of docutils, but a folder in the next. _realpath_dir.intersect calls gen_obj, which calls lstat on /usr/lib/python2.5/site-packages/docutils-0.4-py2.5.egg-info/PKG-INFO, raising OSError, with errno ENOTDIR, but we only catch ENOENT. I'm guessing we should either delete the file, or mark it for replacement, but I'm not sure which or how.