Forwrding to the bug report email as well.
-------- Forwarded Message --------
Subject: Re: Bug#1153978: suricata: 8.0.8 backport conflicts with Rspamd
Date: Mon, 16 Nov 2026 21:15:31 +0200
From: Konstantinos Margaritis <[email protected]>
To: Andreas Dolp <[email protected]>
CC: [email protected], [email protected],
[email protected]
Hi,
First, I have to correct something. The claim of Rspamd maintainer is
based on lack of information, Vectorscan is 100% API/ABI compatible and
I'll leave the quality comment unanswered, I'll just state that we have
cppcheck/clang-tidy in our CI and are running 100s of configurations on
multiple architectures/OSes/compilers, which I doubt is the case with
Hyperscan.
However, I'm planning a new release soon, 5.4.15 -and the respective
Debian package- and I will include the versioned Provides: field then.
This should happen in the next days -I'm actually at SuriCon2026 myself.
That should solve the problem.
Regards
Konstantinos
On 16/11/26 20:52, Andreas Dolp wrote:
Hi Aaron,
thanks for the detailed report. Your analysis matches the package
metadata, but 8.0.7 is not affected, so I am correcting the version
information.
Background: with 1:8.0.8-1 we deliberately switched amd64, arm64,
ppc64el and x32 from Hyperscan to Vectorscan [1], which is the
actively maintained open-source continuation and the one supported by
Suricata upstream. [2]
We do not plan to revert this in unstable.
The conflict is not specific to suricata. It just arises in
combination together with rspamd. libhyperscan5's shlibs file
generates "libhyperscan5 (>= 5.4.2)" in the resulting package, while
libvectorscan5 only carries an unversioned "Provides: libhyperscan5",
which cannot satisfy a versioned dependency (Debian Policy 7.5 [3]).
Both packages ship libhs.so.5, so they cannot be co-installed. Any
package built against Vectorscan is therefore not co-installable with
any package built against Hyperscan, including Debian's own rspamd on
amd64 (on arm64 it already depends on libvectorscan5). I am cloning
this bug to libvectorscan5.
Konstantinos: is the unversioned Provides intentional? Something like
Provides: libhyperscan5 (= ${binary:Version})
would resolve the issue here, because it would fulfill the versioned
Suricata dependency.
Since a versioned Provides would allow rspamd and other reverse
dependencies of libhyperscan5 to run on Vectorscan on amd64, I am
CCing the rspamd and hyperscan maintainers for their view.
Because the upstream rspamd maintainers do not plan to switch to
Vectorscan as stated in [4], I think we should solve this within Debian.
Short term, for trixie-backports: I plan to upload a backport that is
built against Hyperscan again on amd64 and x32, so that the security
fixes can be installed alongside rspamd. For trixie-backports we can
stick on Hyperscan, but please expect this to be changed to Vectorscan
with the upcoming Forky release. I will follow up here.
Regards,
Andreas
[1]
https://salsa.debian.org/pkg-suricata-team/pkg-suricata/-/commit/4ef9c6ad4bc5daa7407d91532f19795ab5ecef45
[2]
https://suricata.io/2026/09/29/why-vectorscan-matters-to-suricata-and-why-were-supporting-it/
[3]
https://www.debian.org/doc/debian-policy/ch-relationships.html#virtual-packages-provides
[4] https://github.com/rspamd/rspamd/issues/6370#issuecomment-6878439420