rmtree problem (again :-( ) on windows

[thomas_mueller_ffb]

Traceback (most recent call last):
  File "C:\Python25\Scripts\bitten-slave-script.py", line 8, in <module>
    load_entry_point('Bitten==0.6dev-r521', 'console_scripts', 'bitten-slave')()

  File "c:\python25\lib\site-packages\Bitten-0.6dev_r521-py2.5.egg\bitten\slave.
py", line 381, in main
    slave.run()
  File "c:\python25\lib\site-packages\Bitten-0.6dev_r521-py2.5.egg\bitten\slave.
py", line 150, in run
    self._create_build()
  File "c:\python25\lib\site-packages\Bitten-0.6dev_r521-py2.5.egg\bitten\slave.
py", line 195, in _create_build
    self._initiate_build(resp.info().get('location'))
  File "c:\python25\lib\site-packages\Bitten-0.6dev_r521-py2.5.egg\bitten\slave.
py", line 207, in _initiate_build
    self._execute_build(build_url, resp)
  File "c:\python25\lib\site-packages\Bitten-0.6dev_r521-py2.5.egg\bitten\slave.
py", line 238, in _execute_build
    _rmtree(basedir)
  File "c:\python25\lib\site-packages\Bitten-0.6dev_r521-py2.5.egg\bitten\slave.
py", line 43, in _rmtree
    return shutil.rmtree(root, onerror=_handle_error)
  File "C:\Python25\lib\shutil.py", line 169, in rmtree
    rmtree(fullname, ignore_errors, onerror)
  File "C:\Python25\lib\shutil.py", line 169, in rmtree
    rmtree(fullname, ignore_errors, onerror)
  File "C:\Python25\lib\shutil.py", line 178, in rmtree
    onerror(os.rmdir, path, sys.exc_info())
  File "c:\python25\lib\site-packages\Bitten-0.6dev_r521-py2.5.egg\bitten\slave.
py", line 42, in _handle_error
    fn(path)
WindowsError: [Error 145] The directory is not empty: 'c:\\docume~1\\deepdi~1\\l
ocals~1\\temp\\bittenufirfj\\build_Qt4.2_6\\deploy\\.svn'

[osimons]

This I've seen before... Subversion makes various files inside a working copy read-only, and shutil.rmtree on Windows won't remove read-only files, and hence directories will not be empty when expected to.

I've got some utility code for another project that wraps rmtree, and if OS is Windows it walks all files and turns off read-only before removing:

import shutil, os
def rmtree(path, *args, **kwargs):
    if os.name == 'nt':
        for (dirpath, dirnames, filenames) in os.walk(path):
            # Windows does not allow remove of read-only files
            for filename in filenames:
                os.chmod(os.path.join(dirpath, filename),
                            stat.S_IWRITE)
    shutil.rmtree(path, *args, **kwargs)

I'll see if I can make a patch to actually test this.

[osimons]

Right. So [521] contains a fix for shutil.rmtree(), and from the traceback it seems you are running that version and still see the error. It seems like that fix sets permissions, but I'm not quite sure that catches a read-only flag on the file in Windows.

Could you please also try to add this line:

bitten/slave.py

@@ -43 +43 @@
     """Catch shutil.rmtree failures on Windows when files are read-only, and only remove if root exists."""
     def _handle_error(fn, path, excinfo):
        os.chmod(path, 0666)
+       os.chmod(path, stat.S_IWRITE)
        fn(path)
     if os.path.exists(root):
         return shutil.rmtree(root, onerror=_handle_error) 

[osimons]

Sorry, forgot the line to include import stat in previous patch. And as the two os.chmod() calls try to do the same it would be interesting just to try the latest:

bitten/slave.py

@@ -20 +20 @@
 import socket
 import tempfile
 import time
+import stat
 
 from bitten.build import BuildError
 from bitten.build.config import Configuration
@@ -42 +43 @@
 def _rmtree(root):
     """Catch shutil.rmtree failures on Windows when files are read-only, and only remove if root exists."""
     def _handle_error(fn, path, excinfo):
-       os.chmod(path, 0666)
+       os.chmod(path, stat.S_IWRITE)
        fn(path)
     if os.path.exists(root):
         return shutil.rmtree(root, onerror=_handle_error)