Deb package
Matthias Andree
matthias.andree at gmx.de
Sat Feb 1 16:18:32 CET 2003
On Fri, 31 Jan 2003, David Relson wrote:
> If memory serves, 0.10.1.4 has only had two problem reports that required a
> code change. The problem was with a message whose payload was of type
> image/bmp and which caused an "Invalid buffer length" message. The fix was
> remove the code to nul terminate the buffer and to start ignoring
> attachments of unknown types (rather than try to decode them). The other
The fix was the NUL termination that made it impossible to satisfy a
read request for 1 byte that was passed down from higher layers. It just
happened to be an image/bmp that triggered the bug, for reasons I
haven't yet full understood, it might be the buffer_fill code that
requests a single byte.
> problem was incorrect order of include statements for gcc.2.7.2.3. The
Euhm, it's not compiler related, the stuff was fine on gcc-2.7.2.3 on my
machine, it's likely a libc problem -- Red Hat 5 might have used libc5
which wasn't on par in terms of a _complete_ implementation...
> other problems have been portability issues in the regression tests - shell
> script incompatibilities and different library treatment of unusual
> characters - specifically of 0xFC which is a "small letter u with
> diaeresis".
It's a common character in German, it appears in "für" (<engl> for) for
example, and depending on what standard the library implementors base
their [[:alpha:]] implementation on, it might be part of this
regexp/ctype or not.
--
Matthias Andree
More information about the Bogofilter
mailing list