Hi Aturo, let me answer your questions inline:
On Tue, 17 Jan 2017 09:48:27 +0100 Arturo Borrero Gonzalez wrote: > Package: hypersan > Severity: wishlist > > Dear hyperscan maintainers, > > thanks for your work with the hyperscan package in debian, it's really > appreciated. > > I just knew that hyperscan v4.4 (which is about to be released by upstream) > will include the so-called 'fat runtime' which enables hyperscan to choose > the instruction set at runtime [0]. > > Upstream states that "there are no changes to how user applications are built > againts the library". That's great. > > From the suricata package point of view we rely on some details of > the current hyperscan debian implementation: > > * the hyperscan install-time warning, which will perhaps no longer be relevant. There is no change in the SSSE3 base dependency with the new 4.4 release as far as I can read from the source code. The fat runtime only ships different optimizations for modern Intel CPUs (AVX2 etc.) and will probably make the overall performance better for most users of the debian package. > * we build twice the suricata binary package, with and w/o hyperscan support. > * we diverge the suricata binary at package install time. > > These things (and some more others) will probably require a spin of adjustement > from our suricata side regarding hyperscan. > > Given the close dates of the stretch release, what do you think about timings? As soon as there is a upstream release I will try to update the hyperscan package promptly and sent you a massage?! > > Also, please, let this bug be a coordination place for the change in > hyperscan vs suricata. > > Feel free to reassign to suricata if you think this bug better fit there. > > [0] https://lists.01.org/pipermail/hyperscan/2017-January/000077.html > > Regards, Rob
