The repr of httplib.IncompleteRead contains all the data in the read so far.

This is stupid. Consider the download of a 100Mb file which fails after
30Mb. You end up with 90Mb of text in the log entry logged with the
python logging framework 
(eg: logger.error('Something bad happened',exc_info=True) )

Would you like to work on a patch?

Please find attached a patch against the trunk.
I'd really appreciate it if this could get merged to the 2.5 and 2.6
branches too, in case of a future release there.

Can't be applied to 2.5 at this point.  I agree it's dumb to
report the entire partial read and that reporting just the
number of bytes read is a much better solution.  Your patch
looks fine to me as well, except you call resp.close() twice
in test_incomplete_read().  Assigning to Benjamin for
application.  (He's going to get the merge stuff right. I
almost certainly will not.)

Why can't it be applied to 2.5?

No problem with the 2nd resp.close() being removed...

The Python 2.5 branch is closed for bug fixes; no further bug fix
releases of Python 2.5 will be made. Only security fixes can be accepted
on the 2.5 branch.

Chris> Why can't it be applied to 2.5?

Benjamin can correct me if I'm wrong, but I thought the last 2.5 release was
the last full release planned.  Certainly if another full 2.5 release is in
the cards then the patch should go there as well.

Skip

Martin> The Python 2.5 branch is closed for bug fixes; no further bug
    Martin> fix releases of Python 2.5 will be made. Only security fixes can
    Martin> be accepted on the 2.5 branch.

So all Chris has to do to get this applied to 2.5 is craft an exploit based
on the current behavior, right? ;-)

S

> So all Chris has to do to get this applied to 2.5 is craft an exploit based
> on the current behavior, right? ;-)

Right :-) Of course, security patches should see a much more careful
review than regular bug fixes.

Applied with a few tweaks in r70107.