[Csv] Test case

[Andrew McNamara]

>    >> I cvs up'd my Python source and confirmed that the problem is fixed
>    >> there.  It's a problem in 2.3 and 2.4 though.  Any chance this can be
>    >> fixed in time for 2.3.5?
>
>    Andrew> The fix involved some radical surgery, so I doubt it's
>    Andrew> appropriate for 2.3.5 - sorry.
>
>Bummer.  

For reference, the parser was partially doing EOL processing in the
line iterator code, partially in the state machine. This meant the EOL
processing had no idea whether it was in a quoted field or not. In 2.5, I
moved all the EOL processing into the state machine. 

>Okay, we have a workaround in SpamBayes, and it is a pretty rare
>corner case for that app.  Since we've bumped into this before do you think
>it warrants a note in the docs?

Possibly. There's a bunch of other stuff of a similar nature that could do
with documenting. I'm inclined to think of it as a bug - while documenting
bugs is a nice thing, there doesn't seem to be much of a precedent for
it in the reference manual.. 8-)

-- 
Andrew McNamara, Senior Developer, Object Craft
http://www.object-craft.com.au/