0 byte read does the wrong thing

[agroce]

import pyfakefs.fake_filesystem
import os

fs = pyfakefs.fake_filesystem.FakeFilesystem()
fs.CreateDirectory('/Volumes')
fs.CreateDirectory('/Volumes/ramdisk')
fs.CreateDirectory('/Volumes/ramdisk/test')
os0 = pyfakefs.fake_filesystem.FakeOsModule(fs)

component0 = "h"
path0 = "/Volumes/ramdisk/test"
path1 = path0 + "/" + component0
fd0 = os0.open(path1, os.O_CREAT | os.O_WRONLY | os.O_TRUNC)
bytes0 = 0
result = os0.read(fd0,bytes0)

should throw OSError with errno 9 (bad fd) according to os

(yes, I added reads just now)

[jmcgeheeiv]

This guy's an animal! Keep going, @agroce!

[agroce]

I'm a lazy professor in the summer, but the random testing tool is indeed an animal. Once it finds something, my job's about three minutes of copy/paste to github. :)

It's running clean on a test regime that omits (of interesting functions in FakeOSModule, e.g. I'm not doing syncs):

1. rename
2. symlinks
3. multiple open file descriptors

since if run with any of those it'll just rediscover the open issues. I expect that #3 is probably hiding a number of fun things once the truncate issues are handled.

I'm also not messing with permissions, since until I get around to dropping this in a virtual machine I can just kill at will, I know from experience that random setting of permission bits in the real file system can turn annoying fast.

[mrbean-bremen]

As I wrote, the low-level file functions (e.g. os.open, os.read, os.write as opposed to the file functions open, read, write) are very likely incomplete, as they are practically not used in application code. I only added basic support for other modes than O_CREAT for os.open a week ago, so there is definitely more to do, but I may handle these with lower priority.

[agroce]

Gotcha. I'll leave the simpler and more straightforward ones at least here for examination, then. I haven't built anything to test open/read/write yet, and aimed to fuzz the underlying FS implementation before the sugar on top, but if open/read/write are doing most of their own work I should handle them at some point.

[mrbean-bremen]

I just understood why the error handling for the low-level functions is incorrect - they are implemented using the high level (fake) functions. This will be a larger rewrite, I leave that for a later time when the temperature is a little down (e.g. compatible with my brain functioning normally).

[mrbean-bremen]

Fixed with #235.