[Rt-devel] BUG: Broken Message eaten up by RT

[Wolfgang Friebel]

On Thu, 24 May 2007, Andrew Sullivan wrote:

> On Thu, May 24, 2007 at 11:06:50AM +0200, Wolfgang Friebel wrote:
>
>> He used pine and stated that the charset he is using is UTF8. The actual
>> mail however does contain umlauts coded in ISO8859-1. The mime attachment
>> announced therefore Text in UTF8 while the content contained chars coded
>> in ISO8859-1.
>
> Hmm.  I can see why you get no body then, because the mailer is
> presumably sending that as an inlined mime attachment (I don't have
> pine installed here to look).  Anyway, from my recentish glance at
> those, if you declare a charset, the other end may try to enforce it,
> and you might get some surprises when the declaration and the data
> don't match.
>
Sure. But if the Charset is in agreement with the content, the message 
looks formally identical (with or without inlined mime) and we do not have 
problems with the ticket contents. Therefore I would expect weird chars 
being transferred to the ticket body in case of disagreement, but not an 
empty or truncated body.

To me it looks like a bug in the UTF8 handling code. In any case I would 
like to see as much text as possible.

BTW it looks like a similar flaw in the program iconv:
If I convert the message with the wrong encoding (it was ISO8859-1)
iconv -f utf8 broken_test_message2

then I get a shortened message and an error message:
iconv: illegal input sequence at position 434

Only if I am using the additional -c flag
iconv -c -f utf8 broken_test_message2

then I do see (almost, a few chars are missing) the full message
-- 
Wolfgang Friebel                   Deutsches Elektronen-Synchrotron DESY
Phone/Fax:  +49 33762 77372/216    Platanenallee 6
Mail: Wolfgang.Friebel AT desy.de  D-15738 Zeuthen  Germany