On Thu, 23 Jul 2026 at 12:45, Eelco Chaudron <[email protected]> wrote: > On 20 Jul 2026, at 15:49, David Marchand via dev wrote: > > > The current pattern for generating string payloads seems a bit heavy. > > Rely on simpler (POSIX) printf syntax using a width format. > > > > Before: > > $ time make -C build check TESTSUITEFLAGS="1135 1136 1261 1262" > > ... > > 1135: dpif-netdev - tso ok > > 1136: dpif-netdev - tunnel tso fallback ok > > ... > > 1261: ofproto-dpif - fragment handling - actions ok > > 1262: ofproto-dpif - fragment handling - reassembly ok > > ... > > real 0m18,511s > > user 0m7,257s > > sys 0m11,074s > > > > After: > > $ time make -C build check TESTSUITEFLAGS="1135 1136 1261 1262" > > ... > > 1135: dpif-netdev - tso ok > > 1136: dpif-netdev - tunnel tso fallback ok > > ... > > 1261: ofproto-dpif - fragment handling - actions ok > > 1262: ofproto-dpif - fragment handling - reassembly ok > > ... > > real 0m5,056s > > user 0m2,807s > > sys 0m2,349s > > > > Signed-off-by: David Marchand <[email protected]> > > Thanks for the fix David. Backported it down to 3.5. For 3.4/3 we might need > a specific patch as the code changed too much.
Thanks Eelco. I worked on this mainly because I added more to the "ofproto-dpif - fragment handling - reassembly" test recently (present in 4.0), and as a consequence, some flow started to expire in some environment. So I don't think we need to backport this further, though it can't hurt. -- David Marchand _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
