bug: "invalid buffer size" with current CVS

[Matthias Andree]

On Fri, 31 Jan 2003, David Relson wrote:

> Matthias,
> 
> Below are the first few lines from bogolexer (verbose/debug mode).  It's a 
> base64 encoded bitmap and function yy_get_next_buffer() in the flex code 
> (not our code) keeps asking for more data until it runs out of buffer space.
> 
> The interesting line in bogolexer's output is "Unknown mime type - 
> 'image/bmp;'".  Seems like at one time we were ignoring attachments.  Seems 
> like that might be the right thing to do for images as bogofilter isn't 
> going to find any meaningful text inside them.  This may take the bulk of 
> the message away from bogofilter's scanning, but bogofilter _does_ still 
> socre the headers for spamicity.
> 
> What's your take on this sitation?

My idea is:

1. feed type text, subtype to the text_html lexer
2. feed type text, any other subtype to the text_plain lexer even if
   that's inaccurate (text/enriched)
3. skip over bodies of any other type.

If we're not doing that at the moment, that's a bug.

However, the bug that caused the bogus abort was the improper shortening
of the buffer by one byte to stuff the NUL character into the end of the
buffer. This is now fixed in CVS.