the function 'email.message_from_file' modifies the message structure when the parsed is invalid (for example, when a closed boudary is missing). The attribute defects is also empty

In the attachment (sample.tgz) you will find:
   - orig.eml : an email with an invalid structure The boundary
"000101020201080900040301" isn't closed
   - after_parsing.eml: same email after calling email.message_from_file()
The boundary is now closed. And the defects attribute is empty
   - test.py: python script to reproduce.

This patch does: 
 - when a close boundary isn't found then the error 
'email.errors.CloseBoundaryNotFoundDefect' is added to the defects list.
 - it doesn't modify the current behaviour of the feedparser 
(eg: the function email.message_from_file still modifies the message 
structure)

with the patch applied: 

{{{
$ ./test.py
PARSER INVALID EMAIL
defects found !
[<email.errors.CloseBoundaryNotFoundDefect instance at 0x7f41421c0488>]
}}}

I also noticed that 'email' modifies the message structure when the header/body separator is missing. And nothing is added to the defect list.
In the attachment, you'll find : 
 - email.patch: this patch add the following error to the defects list :
   - the error 'email.errors.CloseBoundaryNotFoundDefect' when a boundary isn't closed.
   - the error 'email.errors.MissingHeaderBodySeparator' when the header/body isn't found
(patch for python 2.7.2)
 - orig.email: a email without a header/body separator

Thanks for the patch.  I haven't forgotten about it, but it will probably still be a while yet before I get to it.  Hopefully before 3.3 is released, though.

New changeset 81e008f13b4f by R David Murray in branch 'default':
#12515: email now registers a defect if the MIME end boundary is missing.
http://hg.python.org/cpython/rev/81e008f13b4f

I didn't wind up using your patch (for one thing I forgot that there were two separate issues in this patch and independently rediscovered and fixed the MissingHeaderBodySeparatorDefect one).  However, this is now fixed in 3.3.  Unfortunately, since it introduces a new defect, it is an enhancement and by our rules can't be backported.