libera/#devuan-dev/ Friday, 2019-09-06

plasma41the policykit-1 package in beowulf is newer than the version in ceres01:56
fsmithredbut it's probably the same package02:42
onefangIs the owner / admin of dist-mirror.fem.tu-ilmenau.de around?  I'm testing my mirror checker, and I'm getting weird results from them.10:21
onefangWhen my script tries to check for redirects, it gets a http -> https redirect, which works fine, but then https -> https and redirect loop.  Wget and curl are not getting this loop.10:24
onefangThough perhaps wget and curl just ignore a redirect to exactly the same URL?10:25
onefangI don't think wireshark will help, all the important stuff is HTTPS encrypted.10:26
* onefang wanders to the kitchen, it's dinner time here.10:28
onefangYes-, I know it's early Friday morning in .de.10:29
LeePenplasma41: Yes it is the same package just rebuilt for beowulf. Until migration in dak is sorted this is the only way to do it AFAIK.11:19
LeePenIt has since occurred to me that when rebuilding for a different suite, it would make more sense to append ~suiteN (~beowulf1) and then it would show what had gone on more easily and we wouldn't have to do a double version bump for the next unstable upload.11:20
LeePenBut hindsight is a great thing ;)11:20
fsmithredwe could do ~ceres1 and then drop the ~ceres1 for the beowulf version. Same as using ~blah on backports.11:24
LeePenI think that is backwards: doesn't ~ get special sorting in debian package version (*before* an empty string) so 1~beowulf is deemed less than 1?11:30
LeePenSo then we could just increment the next unstable upload by 1 as usual.11:31
fsmithredyes11:31
fsmithredthe beowulf version would still be technically higher than the ceres version, but the part before ~ would be the same number11:32
fsmithredI think gbp does something similar - I've seen versions with ~gbpXXXX11:32
LeePenFor a sane workflow we want the beowulf version <= ceres. If we have 1~ceres1 in ceres and renamed it to 1 for the beowulf build  wouldn't that be beowulf > ceres.11:35
fsmithredyeah, it would. I don't know if there's a way to number them properly.12:02
LeePenI think the other way round works: usual upload to ceres (1), append with tilde for rebuild for beowulf (1~beowulf1). We still have ceres > beowulf.12:32
fsmithredwon't it refuse to build if it's a lower version?12:33
LeePenJenkins?12:35
fsmithredyeah, I think so12:35
LeePenOh, I didn't realise that. Shame.12:35
fsmithredit won't let you build the same version, I'm sure of that.12:36
fsmithredand nobody every suggested bumping the version down to build for beowulf12:36
LeePenI would have thought the version just neede to be different, but I have never tried.12:36
fsmithredneither have I12:37
LeePenWe might just have to continue with the current until the dak migration issue is fixed.12:37
fsmithredanyone know how debian deals with this?12:37
LeePenBut I might have a go when it next arises. It will help make the workflow and version precedence a bit less opaque.12:37
LeePenDebian has britney and migrates in dak AFAIK.12:38
LeePenBut the play the tilde game with backoprts which are versioned ~bpoWhatever12:38
rrqI think it's actually Devuan's "dak" that refuses to upload a rebuild of an existing version.12:39
LeePenrrq: thanks. Helpful. My idea may work then as we will have a different version.12:40
rrqI don't know if there's a problem to build a lesser version than existing ones12:40
rrqI would guess: probably not12:41
LeePenNo worries, it was just an idea as I couldn't sleep in the middle of the night ;)12:42
rrqand your versioning scheme sounds good12:42
rrqthough, maybe it upsets dependency graphs12:42
rrqwe should get the "promotion flow" going12:43
LeePenAbsolutely. Is that Centurion_Dan's baby?12:44
rrqI think I have the power now, and is working on the skill12:45
LeePenLet me know if there is anything I can do to help.12:45
rrqyou speak dak-ish already?12:45
LeePenNo, but I am willing to learn!12:46
rrqsomewhere in the deep depth there's the postgresql database with the meta spaghetti about packages; with a couple of inches of python code for "high(er)-level" actions12:52
rrqsee the devuan-pacakges/dak project, which also has a docs directory with, hmm, documentation12:54
onefangFinally tracked down my problem.  lua-socket supports http, but silently converts http into https, and no mention of that in the docs.12:56
onefangEr *silently converts https into http*.12:59
onefangSo the mirror redirects http -> https, then when my code is checking https lua-socket converts https to http, which the mirror then redirects once more.13:00
fsmithreddoes anyone here have a swap partition? What's in the options field in fstab for it?13:13
fsmithred(I want to know the default setting, not if you changed it.)13:13
rrqdefaults13:16
fsmithredthat was installed with d-i? (not refractainstaller)13:28
fsmithredmy buster has sw in that field, and I can't find it documented (man fstab, man mkswap)13:28
LeePenrrq: where is our dak hosted/running?13:30
rrqhmmm "man swapon" ?13:30
fsmithredyeah, man swapon, not mkswap13:30
fsmithredman fstab refers you to man swapon for swap options. And man swapon refers you to fstab for options.13:31
rrqLeePen: yeah that's the issue.. it's on nextime's VPN13:31
LeePenAnd s/he is no longer around?13:32
rrqfsmithred: those options are for "mount" .. man mount13:32
fsmithredno sw there13:34
rrqnextime is rarely around nowadays, and the VPN access is limited.. he was/is very strict on that point.13:35
fsmithredstrict gatekeeper not at the gate causes trouble for us13:36
rrqand too few people with know-how to set up an alternative13:38
fsmithredI think building a lower version in beowulf will not work. I predict it will complain about the changelog being out of order.13:40
LeePenrrq: thanks for the clarification.13:47
LeePenDo we have a backup if his VPN goes down?13:48
onefangI suspect that's the other half of the problem.13:49
rrqnope.. dan and katolaz had started on the alternative setting up but only half-way so far13:50
rrqonefang: yes13:53
rrqon the balance, that urgency weighs less than keeping it together on the current set up13:56
rrqfsmithred: actually d-i makes fstab say sw for swap .. just tested13:57
rrqapparently from openbsd ?13:58
rrqhttp://man.openbsd.org/fstab13:59
rrqshows that linux dev's are multi-talented, I guess :)13:59
fsmithredI'm trying to decide what to put on that line in refractainstaller. Got a preference? defaults or sw?14:01
fsmithredI think it's been defaults forever. Maybe I should leave it.14:02
rrqwell "man fstab" argues for "none"14:02
fsmithredthat's for the fs field14:02
rrqor maybe I read that wrong, yes14:02
fsmithredI read it four or five times a couple days ago.14:03
rrq"defaults" should be fine... then "mount" does what it does to it14:03
fsmithredI currently have 'swap' in the fs field, and I do need to change it. Nevertheless, it works.14:03
onefangThe real question I should be sorting out is - should mirrors redirecting http -> https be considered a bad idea?  Not every one has apt-transport-https installed.14:09
fsmithredonefang, that would cause failures, wouldn't it?14:12
onefangI assume so, I've not tried it.14:12
fsmithredand redirecting https to http would eliminate the security of https, which some people don't want14:13
fsmithredis there a way to direct https request only to mirrors that support https?14:14
onefangI'm not sure how the system picks which mirror to send people to.  Evilham might know.14:16
rrqambiguous dns resolution14:16
onefangOne problem I'm finding in my tests is Devuan mirror is accessed via https, then redirects to the Debian mirror system, and some of those don't support https.14:18
onefangAnd then my mirror checker gets errors like this -14:19
onefangsslv3 alert handshake failure !   Redirecting to cdn-aws.deb.debian.org -> https://cdn-aws.deb.debian.org/debian/pool/main/d/dash/dash_0.5.8-2.4_amd64.deb14:19
onefangOnce again the weekly mirror status email has lots of errors.  I've been testing my mirror checker script most of the day, and my last run about an hour ago said everything is fine.18:21
onefangDetailed comparison email sent in reply.19:17
Evilhamthere is no "system" in place, deb.devuan.org is a CNAME to a thing that reutnrs a bunch of A records, then the OS / software decides which server to pick at random, standard DNS-based round robin19:40
Evilhamfor completeness: starting with buster https transport for apt is available by default19:41

Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!