Not flushed file gives next results:

rion@rionhost ~/temp $ python2 test.py 
Traceback (most recent call last):
  File "test.py", line 10, in <module>
    m = mmap.mmap(f.fileno(), 0, access=mmap.ACCESS_READ)
mmap.error: [Errno 22] Invalid argument
rion@rionhost ~/temp $ python3 test.py 
Traceback (most recent call last):
  File "test.py", line 10, in <module>
    m = mmap.mmap(f.fileno(), 0, access=mmap.ACCESS_READ)
ValueError: mmap offset is greater than file size


testcase:

import mmap
from tempfile import TemporaryFile

f = TemporaryFile()
f.write(b"hello world")

#f.flush()

m = mmap.mmap(f.fileno(), 0, access=mmap.ACCESS_READ)


just uncomment f.flush() and try again. feel the difference.

suggestion: force flush on file descriptor when passed to mmap.

your suggestion doesn't work. there's no such thing like a "flush on file descriptor".

See also issue #11877 and #11277.

okay guys I'm not going to read all this stuff.
just document it or fix.

thanks for quick reply

@ rion wrote (2011-05-18 12:39+0200):
> just document it or fix.

Hello, zion, Victor, i'm proposing a documentation patch.
It applies to 2.7 and 3.3 (from yesterday).

> Hello, zion, Victor, i'm proposing a documentation patch.

I like it.

I don't think that Python should guess what the user expects (i.e. Python should not sync the file *implicitly*).

@ STINNER Victor wrote (2011-05-18 14:33+0200):
> I don't think that Python should guess what the user expects
> (i.e. Python should not sync the file *implicitly*).

Before i've found F_FULLFSYNC i indeed have had a solution which
tracked the open() of all files, so that mmap() could check wether
a file had been opened in write mode or not.  ;|
Python does not do that by default (AFAIK).

> I don't think that Python should guess what the user expects (i.e. Python should not sync the file *implicitly*).

Agreed.

The documentation patch looks good.

Looked at it again and i think it's much better english with an
additional ..to ensure "that" local...
@Ross, aren't you a native english speaker?  What do you say?

> Looked at it again and i think it's much better english with an
> additional ..to ensure "that" local...
> @Ross, aren't you a native english speaker?  What do you say?

Yes I am, but that doesn't make me any good ;-)

I'd probably agree that it's better with "that" included.

> that doesn't make me any good

Well - 'can only be better than myself, so i'll take that as yes :)

New changeset 198627bbe242 by Ross Lagerwall in branch '3.2':
Issue #12102: Document that buffered files must be flushed before being used
http://hg.python.org/cpython/rev/198627bbe242

New changeset 4898b14dcd69 by Ross Lagerwall in branch 'default':
Issue #12102: Merge with 3.2.
http://hg.python.org/cpython/rev/4898b14dcd69

New changeset e75715f27ce7 by Ross Lagerwall in branch '2.7':
Issue #12102: Document that buffered files must be flushed before being used
http://hg.python.org/cpython/rev/e75715f27ce7

Thanks!