I have been working on lately on a small semantic to the truncate()
method of file objects -- i.e., make file.truncate(pos) imply a seek
to the given argument. I thought I had it all working, but I found out
that one test was failing -- testTruncateOnWindows in test_file. Tried
to fix it by myself, but I failed miserably. So, I am thinking maybe
someone, with a fresh look on the problem, could help me.
By applying the following (simplified) patch:
--- Modules/_fileio.c (revision 59594)
+++ Modules/_fileio.c (working copy)
@@ -635,7 +635,8 @@
return NULL;
}
else {
- Py_INCREF(posobj);
+ /* Move to the position to be truncated. */
+ posobj = portable_lseek(fd, posobj, 0);
}
#if !defined(HAVE_LARGEFILE_SUPPORT)
I get the desired truncate() behavior. However, this causes truncate()
to fail in a weird but predictable manner, as shown by the following
example:
f = open("@test","wb")
f.write(b"1234567890")
f.close()
f = open("@test","rb+")
f.read(4)
f.truncate(4)
print(f.tell()) # should print 4, but print -2 instead (?!)
f.seek(0, 1) # this shouldn't change the file position
print(f.tell()) # print 4 (?!)
It is worthy to note, that calling write(), while tell() returns -2,
raises the following exception:
>>> f.write(b"hello")
Traceback (most recent call last):
...
IOError: [Errno 22] Invalid argument
The thing that I find really weird is that the example translate to
the correct (except for the last write() call) following library
function calls:
f.truncate(4):
lseek64(3, 4, 0, 0, 0xb7da9814) = 4
ftruncate64(3, 4, 0, 0x80aa6c6, 0xb7da9814) = 0
print(f.tell()):
lseek64(3, 0, 0, 1, 0x82ea99c) = 4
write(1, "-2\n", 3) = 3
[The original output of ltrace was much larger; I just show the bare minimum]
There is only three things that comes to my mind that could explain
this weird behavior (in order of likeliness):
1. My code is utter nonsense.
2. I am relying on some undefined behavior.
3. There is a bug in the C library.
Thanks,
-- Alexandre
_______________________________________________
Python-3000 mailing list
Python-3000@python.org
http://mail.python.org/mailman/listinfo/python-3000
Unsubscribe:
http://mail.python.org/mailman/options/python-3000/archive%40mail-archive.com
