Issue1736

os.fstat appears to be in the Jython 2.5.2 documentation, but not in the 2.5.2 os module.

os.fstat is important for avoiding symlink races - a security issue.

AFAIK, Java runtimes offer no direct way of accessing the C libray's fstat(), but fstat remains important for about anything that runs as root and accesses files not owned by root.

We should probably update the documentation.

This bug is also the cause of a failure in our test_httplib.py

http://hg.python.org/jython/file/e80a189574d0/Lib/test/test_httplib.py#l204

The failure of os.fstat(body.fileno()).st_size causes a failure to set a content-length.

The jruby folks have more details about this issue (but they haven't fixed it either)

IO#stat blows up 
http://jira.codehaus.org/browse/JRUBY-3769

Thinking further, we could actually implement a limited version of os.fstat in pure java.

On java 6, we could support size and modification time, through the use of java.io.File. Which is better than nothing, but would not address the requesters issue of dealing with symbolic links.

On java 7, we could implement almost everything, using the new java java.nio.file.Files class, which has extensive facilities, including symbolic link handling, file permissions, etc.

http://docs.oracle.com/javase/7/docs/api/index.html?java/nio/file/Files.html

Thoughts?

I like anything that'll get me st_ino, st_dev, st_uid, st_gid and st_mode.

I'm currently using each of these to detect race condition tampering in a
filesystem backup application.

> "st_ino, st_dev, st_uid, st_gid and st_mode."

On java 7, we can do uid, gid and mode.

But we can't do inode or device, unless there's something I'm missing in the docs.

This is all really low level filesystem stuff that java tries to avoid, because it's so platform specific.

Here is the only mention of inode anywhere in the java docs.

http://docs.oracle.com/javase/7/docs/api/java/nio/file/attribute/BasicFileAttributes.html#fileKey%28%29

"""
Object fileKey()

Returns an object that uniquely identifies the given file, or null if a file key is not available. On some platforms or file systems it is possible to use an identifier, or a combination of identifiers to uniquely identify a file. Such identifiers are important for operations such as file tree traversal in file systems that support symbolic links or file systems that allow a file to be an entry in more than one directory. On UNIX file systems, for example, the device ID and inode are commonly used for such purposes.

The file key returned by this method can only be guaranteed to be unique if the file system and files remain static. Whether a file system re-uses identifiers after a file is deleted is implementation dependent and therefore unspecified.

File keys returned by this method can be compared for equality and are suitable for use in collections. If the file system and files remain static, and two files are the same with non-null file keys, then their file keys are equal.
"""

The existence of fileKey() suggests to me that the os module should have a
new callable that'll detect filesystem race conditions, so that it can be
abstracted as needed in different interpreters.

Naturally, it may never find its way into Python 2.x, but maybe for 3.x.

Fix for 2.7.

Among other things impacts httplib.HTTPConnection send file support, which I'm reexposing as a failure in test_httplib

fstat on file descriptors is supported by JNR Posix (https://github.com/jnr/jnr-posix/blob/master/src/main/java/jnr/posix/LazyPOSIX.java#L124), so we should look at supporting this feature now in 2.7

This lack of support also impacts logging.handlers.WatchedFileHandler

Fixed as of https://hg.python.org/jython/rev/e1a246f6a178