Package: netdiag
Version: 1.0-13
Severity: wishlist
It is illuminating to inject "-Wall" into the makefile
for statnet-3.8, particularly so for the architecture amd64.
The two files
stat_ntohl.c and tally_ntohl.c
are experimenting with hazardous calculations on pointers,
integer type "int", and transforms ntohl()/htonl(). On any
32-bit architecture these calculations are harmless and
even costumary. But for 64-bit architectures they are
indeed questionable.
Installing the package on an amd64 system, and running
the server "statnetd", as well as the client "statnet"
on a single machine, seems to indicate that the compiler
rescues the convoluted code as it stands. However, running
the service for a cluster of mixed endianness, and differing
bit lengths, is a priori another matter. It is to me not at
all certain whether the code is correct.
The upstream changelog even admits that the manipulations
using ntohl() were considered as an untested hack at the time.
That statement was made at a time when 64-bit systems still
were rare.
Thus my remarks might deserve to be kept as a reminder for the
time being, until someone with the proper system resources
can verify the intended functionality, or until the maintainer
grows tired of this issue; whatever happens first!
Best regards,
Mats Erik Andersson, DM
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
