00:01:12 <jclulow> Guest88: So, I'm not sure what those "pkgrepo set" things do
00:01:23 <jclulow> Did someone tell you to do those
00:01:53 <jclulow> Secondly, if you used "nightly" to do the build, "on-nightly" is just the name of the publisher it uses for the packages it creates
00:02:58 <jclulow> It's the default value of "PKGPUBLISHER_REDIST" in Makefile.master
00:03:15 <jclulow> I expect, if you want to have the build use a different publisher, you could override that value in your environment file
00:03:56 <jclulow> What we end up doing for Helios is use the --mog-file option for pkgrecv to apply transformations on the way past
00:05:18 <jclulow> Our tool creates a mogrify file that transforms the package FMRIs to use a different publisher, etc: https://github.com/oxidecomputer/helios/blob/master/tools/helios-build/src/main.rs#L481-L503
00:06:19 <jclulow> Then we transform them when doing a pkgrecv from the nightly*/repo.redist repo into a repo outside the build directory, using --mog-file to point at that file (elsewhere in the program there)
00:20:05 <sommerfeld> jclulow: I can confirm that setting PKGPUBLISHER_REDIST in your nightly environment file works as expected.   
02:12:15 <Guest88> jclulow I found "pkgrepo set" referenced described in a few places. I have been following along and am admittedly a bit shaky on the subcommands. I did use nightly to do the build. Would it be idiomatic to create a pkg/server instance which uses ${path/to/nightly/repo} as its inst_root, thereby serving the new packages via http without the
02:12:16 <Guest88> complication of exchanging packages between local repositories?
02:16:05 <Guest88> Thanks for sharing about Helios. I'm looking at the link now
03:23:04 <jclulow> Guest88: I suppose you could point a depot at the place where the build root lives
03:23:14 <jclulow> You'd probably want to make sure it gets run as the build user/group
03:34:52 <Guest88> Well, coincidentally, I did this (except I had already set it up as root) and it did succeed in transferring the packages, except the publisher name of on-nightly persists even on the other server receiving the packages. I came across some extra info which explains the reason why the publisher name is embedded; apparently, this is intended behavior
03:34:53 <Guest88> (which I hadn't seen documented until now)
04:06:40 <gitomat> [illumos-gate] 13992 VNIC in Zone loses access to properties -- Dan McDonald <danmcd⊙mi>
13:14:22 <gitomat> [illumos-gate] 17148 viona wrongly rejects jumbo frames -- Patrick Mooney <pmooney⊙pc>