-
gitomat
[illumos-gate] 17419 zen_umc logical channel IDs incorrectly determined on some platforms -- Luqman Aden <luqman⊙oc>
-
gitomat
[illumos-gate] 17436 Want option to do broadcast reads with udf -- Luqman Aden <luqman⊙oc>
-
richlowe
sommerfeld: I'm trying an exploratory port of idmapd to your sqlite3 bits
-
richlowe
sommerfeld: one thing I seem to notice is that they have entirely separated the namespace with the '3' suffix, so if you were brave, you could probably link to both
-
sommerfeld
richlowe: yes, I had noticed that and considered that as one way to do conversions.
-
sommerfeld
my general thought is that the way to go is to shift the main consumer from sqlite to sqlite3 and ship a separate converter program (that can eventually be exiled to a separate package that can be uninstalled..)
-
richlowe
that makes sense. It also means if I succeed in this, it would be generally useful
-
richlowe
but that is involving me learning both sqlite APIs, so I wouldn't bet on success.
-
alanc
that's roughly what we did - we shipped 11.4 GA with the binaries (svc.configd, imapd, etc.) linked only against sqlite3 and had a python script we ran before the binaries to read the sqlite2 db and write it back out as sqlite3 (the python script wrapped around the sqlite CLI tool for each version, didn't try using the API)
-
alanc
in 11.4.57 we removed all that and set the minimum version IPS would allow an upgrade from to be 11.4.0, figuring 4.5 years was enough for most of our customers to have upgraded from 11.3, and the rest would just have to live with doing it in two steps
-
jclulow
alanc: How do you express the minimum version for upgrade stuff?
-
alanc
depend type=origin in our system/core-os package
-
jclulow
ah fascinating
-
alanc
which obviously works better when you ensure all your customers get the OS via IPS packages you built, and not a variety of different methods from different distros
-
richlowe
yeah, you'd have to do it at the distribution level, but that's also where it's most useful
-
alanc
specifically, we now have: depend type=origin fmri=pkg:/system/core-os⊙144 root-image=true
-
alanc
(looks like you still call it SUNWcs)
-
richlowe
it's because we're slack, and also can't remember what it was that knew that name
-
alanc
there was some stuff in the slim_source install gate, but I don't remember if there was anything else
-
jclulow
richlowe: I think it's 100% slackness at this point
-
jclulow
witness the slackness at
illumos.org/issues/13625
-
fenix
→
BUG 13625: rename SUNWcs and SUNWcsd (New)
-
alanc
we merged SUNWcsd into other packages, mostly system/kernel as you suggest there
-
alanc
and split out the optional filesystems into their own packages, as Andy suggests there
-
jclulow
Yeah that seems reasonable
-
sommerfeld
who should I ask for bulk builds these days? (want to move 17416/17421 along)