Hello Axel
This compiler flag (-fno-strict-aliasing) is needed in libnids. You are
right. I will revert this change in dsniff. I added the other flag (-g)
in order to avoid creating an empty dbgsym package.
I cannot reproduce this bug in amd64 with the current libnids package
version (1.21). In fact, dsniff works as expected with this testcase.
I built and installed libnids 1.24 (which fix another minor issue
already solved in Ubuntu) with -fno-strict-aliasing. At least in amd64,
dsniff still works as expected and i did not noticed any difference.
If the current maintainer of libnids (CC'ed) gives me permission, i can
import the package to pkg-security team repo, upload my work and
maintain (or co-maintain) it by now. This makes sense as dsniff seems to
be the only reverse dependence of libnids.
I believe that in pkg-security team there are some people with access to
armhf machines so testing should not be an issue. Unfortunately, i can
do it myself.
Greetings,
Marcos
El 17/02/17 a las 00:41, Axel Beckert escribió:
Hi Marcos,
Marcos Fouces wrote:
+dsniff (2.4b1+debian-24) UNRELEASED; urgency=medium
+
+ * Add -fno-strict-aliasing compiler flag in order to fix
+ TCP reassemble in some architectures as armhf.
+ Thanks to guent...@unix-ag.uni-kl.de (Closes: #851060)
+ * Add -g flag to compiler.
So it does not need those compiler flags in libnids but in dsniff?
Have you tested it on armhf?
Because I tried recompiling libnids with these flags and noticed no
difference. :-/
(But then again I could even reproduce the reported issue on amd64, so
I'm not 100% sure if I actually reproduced it correctly.)
Regards, Axel