-
psarriadanmcd, i've uploaded the core of varpd to filepastebin.com, you can download it with this code 054376, mi PI is 20230921T034751Z
-
psarriadanmcd, pastebin.com/AHN5MG9U
-
jperkindanmcd: I'm not actually familiar with the background history to xnet, do you have any good reading material? I'm only familiar with the usual having to set _XOPEN_SOURCE correct to get access to msg_control etc.
-
jperkinhaven't yet come across the case where we've had to add libxnet, so if that's something to look out for I'd like to do it across pkgsrc if possible and add some tests to ensure correct behaviour
-
jperkinto put some numbers on it, currently in 2022Q4 there are 2,382 packages that list a dependency on libsocket, but only 49 on libxnet
-
pjusticeOk, found the behavior change in unbound. This change between 1.17.1 and 1.18.0:
-
pjustice@@ -3809,7 +3860,11 @@
-
pjusticeevbits = UB_EV_READ | UB_EV_PERSIST;
-
pjustice/* ub_event stuff */
-
pjusticec->ev->ev = ub_event_new(base->eb->base, c->fd, evbits,
-
pjustice+#ifdef USE_WINSOCK
-
pjusticecomm_point_udp_callback, c);
-
pjustice+#else
-
pjustice+ comm_point_udp_ancil_callback, c);
-
pjustice+#endif
-
pjusticemeans that where it formerly just used send() or sendto(), it now calls sendmsg().
-
pjusticeif(c->ev->ev == NULL) {
-
pjusticelog_err("could not baseset udp event");
-
pjusticecomm_point_delete(c);
-
pjusticeThat seems broken in and of itself, without getting into the varieties of socket behavior.
-
sjorgehadfl: ^ might be while i am seeing some weirdness with the one from omnios-extra's too
-
sjorgedid not have time to debug so i reverted
-
sjorgeseeing this just reminded me that i was going to try and figure out what broke but forgot
-
pjusticeI tossed a but report into their queue - NLnetLabs/unbound #945
-
pjustice*bug
-
sjorgei suppose in the meantime we can add a patch to revert that one change for both pkgsrc and omnios-extra
-
pjusticeUSE_WINSOCK wouldn't seem to be particularly relevant for either of those platforms.
-
pjustice;)