Message258

This is related to a bug I submitted a month or so ago.  You guys rightfully closed it, but I finally nailed down what was causing it.  There should be at least be a note somewhere in the documentation about it.  I'm not sure it can be fixed.

I know, technically, you should always close your files.  But what if you don't?  In CPython, if you don't close your files when opening them in a function, the garbage collector closes it for you.  But there has been some inconsisencies in how Jython behaves.

Here is my test:


import os,tempfile

def writeFile(fileName,str): 
	file = open(fileName,"w")

	file.write(str)
	# I'm not closing the file on purpose

if __name__ == '__main__':

	fileName = "bugtest.tmp"
	str = "Frank Burns eats worms\n"
	writeFile(fileName,str)

	os.remove(fileName)

This works fine on CPython on Windows 2000 and Solaris as well as Jython 2.0 on Solaris, but not with Jython 2.0 on Windows 2000.  On Windows 2000, it gives this exception:

Traceback (innermost last):
  File "bug.py", line 16, in ?
  File "C:\jython-2.0\Lib\javaos.py", line 44, in remove
OSError: [Errno 0] couldn't delete file: 'bugtest.tmp'

In reality, I think the behavior on Windows 2000 is the correct one, but the choice is up to you.  We did some research here, and this really does go down to the C level.  Following is an example, which we compiled with gcc on Solaris, gcc on Cygwin, and lc (the line compiler with Visual Studio).

#include <stdio.h>

int main() {
  FILE *f;
  f=fopen("tempfile","r");
  if(f != NULL) {
     printf("File is Open\nAttempting to Delete...");
  }
  if(remove("tempfile")!=0) {
    printf("Delete Failed\n");
  } else {
    printf("Delete Successful (Solaris Sucks)\n");
  }
}

With both versions of gcc, this file was deleted.  With lc, the delete failed.  (Obviously, the file "tempfile" needs to be created before running this).