<@nirik:matrix.scrye.com>
16:00:06
!startmeeting riscv-sig
<@meetbot:fedora.im>
16:00:08
Meeting started at 2025-03-04 16:00:06 UTC
<@meetbot:fedora.im>
16:00:08
The Meeting name is 'riscv-sig'
<@nirik:matrix.scrye.com>
16:00:10
!topic init process / agenda gathering
<@nirik:matrix.scrye.com>
16:00:10
!meetingname riscv-sig
<@meetbot:fedora.im>
16:00:11
The Meeting Name is now riscv-sig
<@davidlt:matrix.org>
16:01:23
!hi
<@zodbot:fedora.im>
16:01:24
David Abdurachmanov (davidlt)
<@abologna:matrix.org>
16:01:30
hello
<@nirik:matrix.scrye.com>
16:01:47
morning everyone
<@nhanlon:beeper.com>
16:01:52
!hi
<@zodbot:fedora.im>
16:01:53
Neil Hanlon (neil) - he / him / his
<@nhanlon:beeper.com>
16:01:55
g'day folks
<@nirik:matrix.scrye.com>
16:03:05
So, I had a few infra things: I have keytabs working with the secondary hub, so we can look at onboarding builders.
<@nirik:matrix.scrye.com>
16:03:24
also added some folks with perms to avoid me being a blocker. ;)
<@rwmj:matrix.org>
16:03:25
'lo
<@davidlt:matrix.org>
16:03:51
nirik: do you have full instructions somewhere to try this?
<@nirik:matrix.scrye.com>
16:03:55
I think abologna was going to look at getting a builder with the new image online / configured / documented.
<@rwmj:matrix.org>
16:04:02
davidlt: should we move builders over yet? I think last time you said not to do it yet
<@abologna:matrix.org>
16:04:29
I've only gotten as far as flashing a blank image on an SD card so far :'(
<@davidlt:matrix.org>
16:04:41
No, if we move the builders now everything stops. We prepare and switch when everything is ready.
<@nirik:matrix.scrye.com>
16:04:42
I don't, but they are pretty obvious... get the keytab, set kojid.conf to use it, set hostname to match it, set koji url for secondary hub, profit. ;)
<@abologna:matrix.org>
16:05:01
nirik: maybe not so obvious for someone who's never done it before ;)
<@davidlt:matrix.org>
16:05:16
All good. Let's wait for abologna to deliver instructions and his experience 😄
<@rwmj:matrix.org>
16:05:45
ok
<@abologna:matrix.org>
16:05:55
I can promise to make at least *some* progress over the next couple of days. then I'm away until next Tuesday
<@davidlt:matrix.org>
16:06:00
I will give a try once abologna has some instructions.
<@nirik:matrix.scrye.com>
16:06:23
Then, the next big topic for today is the overlay again I think. ;)
<@abologna:matrix.org>
16:06:37
oh boy :)
<@davidlt:matrix.org>
16:07:25
nirik: I said this multiple time probably, but I am not the one setting it up. I am flexible to the point where whatever you pick probably works for me 😄
<@nirik:matrix.scrye.com>
16:08:03
well, would it be possible to just keep using the gitea one you have? or thats bad because only you can commit to it?
<@davidlt:matrix.org>
16:08:30
Doesn't work. We are slowly breaking apart. kernel-ark is killing Gitea (a bug, as it used to work).
<@nirik:matrix.scrye.com>
16:08:47
kernel-arc is a beast of a git repo indeed.
<@davidlt:matrix.org>
16:08:54
I have restricted resource usage (via systemd) otherwise it starts killing Koji Hub, Web, etc.
<@nirik:matrix.scrye.com>
16:09:08
icky. ok.
<@davidlt:matrix.org>
16:09:11
It's not integrated with FAS too, which would be nice to have.
<@abologna:matrix.org>
16:09:23
I'm still leaning towards a group on pagure.io
<@rwmj:matrix.org>
16:09:40
just because it's a huge git repo?
<@abologna:matrix.org>
16:09:49
but I'm thinking, can we maybe do some actual tests once we have at least one builder up and running?
<@davidlt:matrix.org>
16:10:02
I am fine with whatever nirik says we can do. My biggest goal here is to have Fedora / FAS integration and make things simple for Fedora Infra.
<@rwmj:matrix.org>
16:10:13
on my copy of kernel-ark
<@rwmj:matrix.org>
16:10:16
3.7G .git
<@rwmj:matrix.org>
16:10:16
$ du -sh .git
<@davidlt:matrix.org>
16:10:29
Richard Jones: some indexing, health check, etc. doesn't run efficiently on this repo.
<@abologna:matrix.org>
16:10:35
yeah FAS support would be amazing
<@davidlt:matrix.org>
16:10:46
It literally almost went to <300 on load averages one day.
<@nirik:matrix.scrye.com>
16:10:46
I don't want to spend a lot of cycles setting this up as we will just be trying to get rid of it anyhow. ;) So it's just short term...
<@nirik:matrix.scrye.com>
16:11:08
so I guess from my point of view pagure would be easiest.
<@abologna:matrix.org>
16:11:10
pagure.io and gitlab.com both have fas support and good capabilities for collaboration
<@abologna:matrix.org>
16:11:20
the latter is problematic from the firewall point of view IIUC
<@nirik:matrix.scrye.com>
16:11:28
gitlab is anoying due to the load balancer issue. ;( yeah.
<@davidlt:matrix.org>
16:11:32
nirik: not exactly, every time we need to bump (i.e. rebuild) a package it also lands in dist-git overlay. We don't want to bump the official dist-git.
<@abologna:matrix.org>
16:11:54
the former is a bit unwieldy and some{what,times} slow
<@davidlt:matrix.org>
16:12:17
Gitlab uses Google, Cloudflare (CDN), Amazon, etc. It's a long list IP ranges.
<@nhanlon:beeper.com>
16:12:39
on the other hand we mostly all work with and love (?) pagure despite it's faults on a daily-ish basis
<@abologna:matrix.org>
16:12:39
IIUC the list is big but well-defined. doesn't that help?
<@davidlt:matrix.org>
16:12:56
It's not. The IP I get from Cloudflare is not in the IP ranges provided.
<@nirik:matrix.scrye.com>
16:12:57
its not well defined that I could find. ;)
<@abologna:matrix.org>
16:13:16
I believe davidlt had found a page with all the ranges listed
<@rwmj:matrix.org>
16:13:35
pagure isn't the worst tool i have to use, but not the best either
<@nirik:matrix.scrye.com>
16:13:41
so I could do the thing I did for osbuild, but it's anoying. It's a cron that runs every minute and updates a ipset with the current ips.
<@davidlt:matrix.org>
16:13:43
Yes, but IP I get is not within it. Yet Internet knows it's a dedicated IP in Cloudflare network for Gitlab.
<@abologna:matrix.org>
16:14:26
regarding rebuilds. are those mostly needed for bootstraps and so on?
<@abologna:matrix.org>
16:14:48
could we use side tags? reduce the number of commits that we need to do outside of dist-git somehow?
<@davidlt:matrix.org>
16:14:52
Also, we would be passing full Cloudflare through the firewall. That's a large chunk of Internet itself these days.
<@davidlt:matrix.org>
16:15:17
Side-tags don't solve this problem as you still need to bump the SPEC file.
<@nirik:matrix.scrye.com>
16:15:43
I wonder if we could look at how eln works there. :) They use a dist tag... and bump that dist tag whenever they need to rebuild...
<@davidlt:matrix.org>
16:16:01
It's for bootstrap, but mostly fixing packages to build with proper dependencies (e.g. SOVERSION bumps), and in some cases rebuilds due to a toolchain bug.
<@abologna:matrix.org>
16:16:07
but if most are bootstrap commits that means there's going to be a later commit effectively reverting them later, right?
<@nhanlon:beeper.com>
16:16:14
_barely suppresses a rant about centralization of internet resources_
<@davidlt:matrix.org>
16:16:49
We don't revert. We use a different NVR pattern today, and the script just ignores it. If there a newer one in upstream it picks it up.
<@abologna:matrix.org>
16:17:17
right, I'm thinking in terms of the future where we want to be closer to Fedora is possible
<@nirik:matrix.scrye.com>
16:17:37
https://docs.fedoraproject.org/en-US/packaging-guidelines/#bootstrapping
<@davidlt:matrix.org>
16:17:53
It's just a rebuild. Tiny SPEC change which is ignored by my scripts, and vanishes as soon as there is newer NVR in Fedora.
<@nirik:matrix.scrye.com>
16:18:10
huh, thats the old doc. can't find the new one.
<@abologna:matrix.org>
16:19:29
let's table this discussion for now, the details are effectively irrelevant, the basic point is that we're going to need the overlay for a while still unfortunately
<@nirik:matrix.scrye.com>
16:19:34
basically you can do it in a side tag, set the _bootstrap macro, build, unset it, build again and only one bump to make the orig change in version or whatever
<@nirik:matrix.scrye.com>
16:19:41
yeah, getting sidetracked.
<@abologna:matrix.org>
16:19:59
we can improve our process over time to reduce the need to a minimum
<@abologna:matrix.org>
16:20:28
but the need is never going away until we're on the primary koji I think
<@nirik:matrix.scrye.com>
16:20:29
so, lets look at this once we have a builder and see what we can work with easiest?
<@nirik:matrix.scrye.com>
16:20:53
there's also codeberg I guess (not sure their setup for ips)
<@davidlt:matrix.org>
16:21:06
`_bootstrap` is not what we need.
<@abologna:matrix.org>
16:21:11
I'm not familiar with it, does it gave FAS integration?
<@abologna:matrix.org>
16:21:26
if not that would make it lower on the list than pagure.io IMO
<@nirik:matrix.scrye.com>
16:21:41
I don't think yet... it's it's own accounts.
<@nirik:matrix.scrye.com>
16:22:08
so, how about we try and make pagure work then? If something doesn't work there, we revisit?
<@abologna:matrix.org>
16:22:13
I really like the idea of gradually getting closer to Fedora infra. having both the dist-git overlay and our own koji with FAS integration would be a great step in that direction
<@davidlt:matrix.org>
16:22:54
FYI: https://pagure.io/koji/pull-request/898#
<@davidlt:matrix.org>
16:23:18
I see side tag allows settings macros too. I wonder if this allows overriding dist macro.
<@nirik:matrix.scrye.com>
16:23:45
in primary, no. in secondary... we could allow it
<@davidlt:matrix.org>
16:24:29
I still don't like it. Bumping SPEC is traditional way.
<@nirik:matrix.scrye.com>
16:25:05
well, or a comment only git commit for rpmautospec using things. ;)
<@davidlt:matrix.org>
16:26:17
Not really, because we cannot take upstream NVRs. We pass `-e rvreX` to autorelease macro to avoid this.
<@nirik:matrix.scrye.com>
16:26:54
ok. we probibly should document this stuff? ;) (Or perhaps it is?)
<@davidlt:matrix.org>
16:26:58
Anything we touch we make sure future NVRs from upstream are not affected.
<@davidlt:matrix.org>
16:27:35
It's not on wiki page, but it was documented somewhere, and mentioned multiple times a year for years.
<@davidlt:matrix.org>
16:28:40
Anyways, we need dist-git repo to do these stuff, and it should have API too.
<@davidlt:matrix.org>
16:28:49
(because it's not done manually)
<@nirik:matrix.scrye.com>
16:28:58
ok. I'm not saying it's bad or anything, I just don't know how you do things day to day. ;)
<@nirik:matrix.scrye.com>
16:29:26
I think we should just try pagure and if it's not workable for some reason we can revisit.
<@abologna:matrix.org>
16:30:27
I'm personally on board with that plan
<@davidlt:matrix.org>
16:31:36
I am not finding any Python packages for Pagure and it's API.
<@abologna:matrix.org>
16:32:11
davidlt: it's `python3-libpagure`
<@abologna:matrix.org>
16:32:23
I'm using it in the tracker, to fetch information about pull requests
<@davidlt:matrix.org>
16:32:36
REST + JSON for Pagure is an option too.
<@abologna:matrix.org>
16:32:39
very minimal use so I won't comment on its quality
<@davidlt:matrix.org>
16:32:56
libpagure seems to be dead, last update 5-7 years ago
<@abologna:matrix.org>
16:33:45
that's a large time range :D
<@abologna:matrix.org>
16:34:08
if the API hasn't changed much in the meantime there might not be a reason to update it
<@abologna:matrix.org>
16:34:25
I don't think pagure itself has seen a lot of development in recent years
<@abologna:matrix.org>
16:35:00
libpagure certainly works when used against the src.fedoraproject.org instance, at least for the operations I've tried
<@abologna:matrix.org>
16:35:22
I have no reason to believe it wouldn't work against the pagure.io instance
<@smooge:fedora.im>
16:36:05
davidlt: yes that is about the time it has had major work done to it.
<@davidlt:matrix.org>
16:36:14
I am looking into APIs right now.
<@abologna:matrix.org>
16:38:12
do we have other topics for the meeting? otherwise I think this is something that we can continue discussing in the regular channel
<@abologna:matrix.org>
16:38:23
and we can cut the meeting short
<@davidlt:matrix.org>
16:38:43
Container registry.
<@davidlt:matrix.org>
16:39:24
I think we can use what nirik suggested.
<@nirik:matrix.scrye.com>
16:39:29
ok, back
<@nirik:matrix.scrye.com>
16:39:50
yeah, if we have containers we could upload to the fedora-testing org on quay.io...
<@nirik:matrix.scrye.com>
16:40:12
I wonder if the emulation works right there to run those on a x86_64 machine...
<@davidlt:matrix.org>
16:40:15
Do we need to think of a name? fedora-riscv?
<@davidlt:matrix.org>
16:40:54
You should be able to run non-native container if you setup podman/docker with qemu stuff.
<@nirik:matrix.scrye.com>
16:41:00
sure, seems reasonable
<@jmontleon:fedora.im>
16:41:09
I run riscv containers with qemu-user-static on x86_64 a lot if that's what you mean
<@nirik:matrix.scrye.com>
16:41:13
how often do we make images?
<@davidlt:matrix.org>
16:41:23
Pagure API does not seem to nice after spending initial 5+ minutes looking at it.
<@nirik:matrix.scrye.com>
16:41:25
Jason Montleon yeah... cool.
<@davidlt:matrix.org>
16:41:40
Not often, but we will increase timing.
<@davidlt:matrix.org>
16:41:47
I would love to go back to daily images at some point.
<@nirik:matrix.scrye.com>
16:42:00
yeah, would be nice.
<@abologna:matrix.org>
16:42:10
that seems somewhat unnecessary, but sure
<@davidlt:matrix.org>
16:42:12
Current images are stable snapshots, and we decided when to make them.
<@nirik:matrix.scrye.com>
16:42:30
I guess for now, I can setup a robot account and get the creds to davidlt and he can upload them? or ?
<@abologna:matrix.org>
16:42:33
we need at least to match disk images
<@davidlt:matrix.org>
16:42:53
and at least to abologna
<@davidlt:matrix.org>
16:43:03
He cooked the last images 😄
<@abologna:matrix.org>
16:43:12
remind me again, where would these image live?
<@nirik:matrix.scrye.com>
16:43:28
https://quay.io/organization/fedora-testing
<@jmontleon:fedora.im>
16:44:04
its does NOT work on ppc64le; buts that's just page size if I recall; don't know if that would stop some of the arm stuff/system from working (i think that people are slightly more likely to try this)
<@davidlt:matrix.org>
16:44:10
Currently OCI tarballs and fedora.riscv.rocks registry is where they are placed, which is not secure registry (http).
<@nirik:matrix.scrye.com>
16:44:30
once we have something there we should make some noise about it... so people know it's there/can test it.
<@nirik:matrix.scrye.com>
16:44:56
anyhow, can get that setup out of meeting I suppose too
<@davidlt:matrix.org>
16:45:01
Sounds like wiki + Fedora Magazie article on RISCV and how to run non-native containers 😄
<@davidlt:matrix.org>
16:45:11
It's like 1-2 lines before executing the container.
<@abologna:matrix.org>
16:45:17
yeah that would be great! similar to the existing QEMU images
<@abologna:matrix.org>
16:45:22
*instructions
<@abologna:matrix.org>
16:45:43
give people a way to conveniently try out Fedora RISC-V without hardware purchase
<@nirik:matrix.scrye.com>
16:45:54
yep 💯
<@jmontleon:fedora.im>
16:45:59
ya, basically dnf install qemu-user-static-riscv podman or whatever the combo is and go
<@davidlt:matrix.org>
16:46:04
Having images on Quay means that it's a secure registry, and it's configured by default. Note, that toolbox configuration still needs to be modified to work.
<@nirik:matrix.scrye.com>
16:46:41
I suppose package installs in there would pull from koji repos currently? or ?
<@abologna:matrix.org>
16:46:55
yeah, same as the disk images
<@abologna:matrix.org>
16:47:15
we have a patched `fedora-repos` package that points there and is identical between disk and container images
<@nirik:matrix.scrye.com>
16:47:18
thats fine for now.
<@abologna:matrix.org>
16:47:37
more like no other way for now :)
<@davidlt:matrix.org>
16:47:47
We used to pack podman and toolbox configuration, but it was removed from the Kiwi and replaced with wiki instructions.
<@nirik:matrix.scrye.com>
16:48:00
so, about ~13m left... any other topics?
<@rwmj:matrix.org>
16:48:45
not from me
<@abologna:matrix.org>
16:48:45
not from my side. we can discuss the details in the regular channel
<@nirik:matrix.scrye.com>
16:49:30
sounds good... will close out in a min then?
<@davidlt:matrix.org>
16:49:41
ok
<@nirik:matrix.scrye.com>
16:50:18
ok, thanks for coming everyone, lets continue in #riscv:fedoraproject.org
<@nirik:matrix.scrye.com>
16:50:21
!endmeeting