[imp] Base64-only message not quoted in reply

[Robin Bankhead]

Quoting Robin Bankhead <horde at headbank.co.uk>:

> Quoting Michael M Slusarz <slusarz at horde.org>:
>
>> Quoting Robin Bankhead <horde at headbank.co.uk>:
>>
>>> Quoting Robin Bankhead <horde at headbank.co.uk>:
>>>
>>>> Quoting Michael M Slusarz <slusarz at horde.org>:
>>>>
>>>>> Quoting Robin Bankhead <horde at headbank.co.uk>:
>>>>>
>>>>>> Didn't want to ticket this automatically in case it's put down  
>>>>>> to bad message format (I'm no expert but this seems a common  
>>>>>> reply to such issues).
>>>>>>
>>>>>> The (redacted) message below is all base64-encoded text/plain  
>>>>>> and has no plaintext part.  It views fine, but when replying to  
>>>>>> it, it is not quoted.  We have no issues with the original HTML  
>>>>>> + quoted-printable messages from the same sender, but their  
>>>>>> replies to our replies always come in this base64-only format  
>>>>>> (I wonder if their Barracuda spam firewall is doing this?)
>>>>>
>>>>> This *faintly* sounds like an issue that may have occurred many  
>>>>> years ago, but IIRC it was either an issue with a specific PHP  
>>>>> version (possibly dealing with variable references) or has long  
>>>>> since been fixed in IMP and/or Horde.
>>>>>
>>>>> michael
>>>>>
>>>> Indeed, I was going off this old ticket[1], though the difference  
>>>> is that person got raw base64 in the reply quote; I get nothing.
>>>>
>>>> PHP version is 4.3.1; currently I can (if necessary) upgrade to a  
>>>> later 4.3 version, but I'm not able to go beyond that without  
>>>> full code-review of the other application on the same server.
>>>>
>>>> Horde platform is groupware-H5 pear on Windows/Apache2.2, all up  
>>>> to date excepting imp-6.1.5 (and I see nothing in changelog that  
>>>> might address this).
>>>>
>>>> Thanks,
>>>> Robin Bankhead
>>>>
>>> Arggg - sorry, the above should have read PHP 5.3.1 (Thanks for  
>>> the catch, Simon).
>>
>> Not sure what ticket you are linking to... there was no link in the message.
>>
> Sorry, forgot to footnote it:
>
> http://bugs.horde.org/ticket/?id=1077
>
> As you say, very old; different symptoms though.
>
>> I *really* think this is probably the issue though.  My memory is  
>> that there were bugs in the past relating to PHP variable  
>> references that may cause this kind of behavior.  And I would also  
>> classify PHP 5.3.1 as ancient ... it's over 5 years old.  I  
>> personally would not want to be running 5 year old code that is  
>> potentially publicly accessible to the Internet, if just for  
>> security reasons.
>>
>> michael
>>
> Personally I wouldn't either, but it's not, so I have the luxury of  
> prioritising the stability of my codebase.  Even so, I bet you'd  
> find plenty of web hosts where 5.3 is still deployed - the need to  
> guarantee *functional* continuity/stability is pretty big in that  
> context too. I rather imagined that was why horde's INSTALL file  
> specifies the requirement as 5.3.0 and up.
>
> Could you offer any pointers as to where in the code I should be looking?
>
> Regards,
> Robin Bankhead
>

Update: it appears base64 is a red herring, as I've found other such  
messages in our corpus that don't exhibit this problem.  I will  
compare the messages in detail to try to work out how else they differ.

Could the Accept-Language or Content-Language headers  have any effect  
here?  I really don't see any other notable differences other than the  
non-standard ("X-*") headers, which I assume IMP has no interest in.   
The "Good" and "Bad" examples are both sent by MS Exchange so it's  
only the antispam headers that differ.

Regards,
Robin Bankhead