Wednesday, 2023-11-08

JayFtc-members: can I get at least one ack on this before I pop it out to the mailing list?
JayFwhoops, ECHAN00:04
tonybLOL, I didn't even notice00:05
JayFwell, I notice it's my EOD, I got that done just in time to leave, finally getting to the end of my PTG documentation duties :D 00:07
JayFdid the ironic spec and the TC email today, assuming nobody from either end nacks it :D 00:07
JayF (Ironic does it's PTG documentation as a spec)00:07
tonybInteresting idea00:08
JayF I can't take credit, we've been doing it since mitaka00:08
JayFjust changed the name from priorities -> themes -> work items over time00:08
tonybOh nice00:09
JayFtrying to figure out whatever word we can that says "this is our wishlist of stuff we'd like to work on, but the real world existing means we'll get like, half of it done if we're lucky" :D 00:09
JayFit's extremely satisfying to be able to go back in time and see the things that you've done though00:09
clarkbI've been nerding out on unicode/utf8 because I got nerd sniped. Its interesting that the vim 'ga' command shows me the unicode point and not the utf8value even though the file is encoded in utf800:09
clarkbhexdumping confirms.00:09
JayFclarkb: we actually have code that deals in UTF-16 in ironic (IPA!)00:10
JayFclarkb: fallback uefi bootloader uses csv files that are encoded UTF-1600:10
JayFor was it -3200:10
JayFnow I gotta go look00:10
opendevreviewClark Boylan proposed opendev/system-config master: A file with tab delimited useful utf8 chars
JayFyep, utf-16 or utf-16-le00:11
JayFendianness matters, yikes, I forgot about that00:11
* JayF out for the day o/00:11
clarkbother than failing to properly indicate the tabs in all cases that actually rendered well00:12
clarkbI'm not actually sure we want to merge a change like that. However it might be useful to have four byte utf8 chars readily available somewhere like that due to all the historical problems with mysql and those characters00:13
clarkbalso seriously everything has a code point. Even symbols from clay tablets00:14
tonybHuh, I can see advantages to having something like that.  How much do you trust the pre-deploy testing ;P00:16
clarkbtonyb: so far its just adding the content to he git repo which should be pretty safe :)00:17
clarkbbut the background idea here is maybe trying to add a testinfra test to etherpad that uses the etherpad admin api to inject 4 byte character pad data then eftches it back out again00:17
clarkbthe first step was go figure out what even are four byte characters and then I ended up deep down the rabbit hole of wow there are a lot of characters00:17
clarkband kind of forgot the original idea here :)00:18
tonybIt happens00:25
tonybI started today with the goal of proposing gnocchi to g-r and now I'm rewriting our constraints generation infrastructure 00:29
yadneshfungi, at times when downloading fedora images from official fedora urls in telemetry ci fips job fails with error "GnuTLS: One of the involved algorithms has insufficient security level." This happens with some specific mirrors. , openstack04:30
yadneshtelemetry team was planning to switch to a different image source, would it be possible to sync images to opendev mirrors ?04:30
fricklergtema: is still relevant or can it be abandoned? 07:48
* frickler doing a bit of backlog trimming07:49
gtemauhm, honestly speaking I do not know. I changed company, but this still can be done for the glory of OpenStack ;-)07:50
gtemagimme maybe half a day and I will try to check with various parties whether this makes sense (it does make sense, but it requires some dedication)07:52
fricklergtema: no more otc?08:00
gtemaI am not there anymore08:00
eanderssonApparently there is a configuration option for coverage that fixes coverage issues with greenlet.11:17
fricklereandersson: seeing your followup in #-dns it looks like you were just talking to yourself? ;)12:16
eanderssonYea just justifying the two useless rebases12:29
Clark[m]yadnesh: we stopped building and mirroring fedora because it's cadence is just too quick for us to keep up. For example the fedora 37 image your job uses is end of life in less than a week. My suggestion would be that you report the broken mirrors to fedora so they can be fixed and possibly look at alternative images that don't EOL every 13 months.13:54
*** amoralej_ is now known as amoralej14:06
fungiit sounds like the desire is for a more full-featured image to boot in devstack, as opposed to cirros?14:13
ildikovHi OpenDev Team, I have a patch in the StarlingX election repo that is approved, but didn't reach the merged status: Can you help me figure out what went wrong with it?14:15
fungiildikov: its parent is abandoned14:16
fungirebase it onto the current branch state rather than on top of the abandoned change14:17
ildikovfungi: good catch, thank you!14:17
fungiany time!14:18
ildikovI think I stared at it too long to see that trivial issue, lol14:18
fungiClark[m]: aha, i knew if i was too lazy to look into the compress error, someone else would report it upstream anyway! something about sitting on the river bank while the corpses of your enemies float by...
tonybyadnesh: You should for sure report that to the Fedora mirror admins, meanwhile if you can provide details on what you're looking for by using Fedora that'd be helpful14:20
Clark[m]fungi: looks like they closed it. The changelog entry there probably makes more sense to someone that understands what that library does than it does to me 14:25
ildikovfungi: I rebased the change and Gerrit shows that it's in Work in Progress state. How can I make it 'Active'?14:26
fungiClark[m]: i think it's that mailman needs to add an extra template file14:26
ildikovfungi: I made edits on it to resolve the merge conflict14:28
fungiildikov: doesn't seem to allow anyone other than the change owner to toggle wip state14:29
ildikovfungi: isn't that the workflow label?14:30
tonybNope Worflow is +W14:30
tonybWiP is different14:30
ildikovtonyb: ah, ok, because W -1 puts the patch into 'a' WIP state14:31
ildikovwhich is where my confusion is coming from with this one14:31
ildikovhow can the patch owner fix this?14:32
tonybThey could/should rebase it onto the current master of the election repo and if it's still a "work in progress" they can filp that under the "3 dots" menu to the far right of the gerrit UI14:33
tonybildikov: If starlingX's election process doesn't require the change owner and the candidate to be the same account then I/you could do that14:34
fungior adjust the acl for that repo to allow core reviewers to toggle wip on changes too14:35
tonybfungi: That too14:35
fungian example:
ildikovtonyb: the patch owner needed to fix the commit message and they did it as a new commit, but I missed that the new commit was on top of the original one that needed to be fixed. So I abandoned the orginal one already. Then when the issue got clear to me, then I went ahead and rebased the change to fix it, but it ended up in a weird merge conflict, which I resolved, but couldn't remove the patch from WIP.14:36
fungia bit of history, workflow -1 votes were our old workaround for gerrit's lack of a wip state. then more recently, gerrit finally added a built-in wip feature, but acls for projects haven't really caught up14:37
ildikovtonyb: The StarlingX election process doesn't say anything about the details of how to fix issues like this, and that the owner has to.14:37
ildikovfungi: ah, thanks for sharing that detail, I did miss that too, regarding Gerrit's 'WIP' state14:38
ildikovtonyb: fungi: many times election officials or someone else help the submitter with submitting their noinations, so having either the election officials or you fix this current issue would be perfectly fine14:39
tonybildikov: Okay I'll fix it ... gimme 514:39
ildikovfungi: and I think having cores be able to toggle WIP would also be a good way forward14:39
fungia bunch of openstack projects have added something like "toggleWipState = group myproject-core" to their acls, the starlingx community could consider doing the same14:39
ildikovfungi: thank you!!14:39
ildikovfungi: I'll share that on the ML14:39
tonybildikov: just be 100% clear we're talking about
fungiotherwise, yes, you can get in a situation where some change owner sets wip on their change, then goes on vacation and the core reviewers can't merge it until they return14:40
fungisince wip state is "sticky" and persists into subsequent revisions until it gets unset explicitly14:41
Clark[m]Note, gerrit doesn't seem to have recorded a switch to wip so I think it was pushed this way. I would also encourage people to not push code as wip unless they don't want it to be merged14:41
Clark[m]In this case it seems the intent is to make it mergable so wip is counter productive14:42
tonybclarkb: I think it landed that way because of the mergeconflict and parent issues14:42
fungiwell, gerrit won't automatically set work-in-progress on changes14:43
fungigit-review has a -w option where you can set wip when pushing a change though14:43
Clark[m]That doesn't put things into a wip state though? You have to either explicitly select the state in the UI or explicitly push to the WIP push destination 14:43
Clark[m]I'm saying people should be told to only do that if their intention is something not merge 14:44
Clark[m]In this case the intention is to merge a candidacy so wip is problematic 14:44
fungii've seen this before with starlingx changes. i gave someone feedback on a change which was in wip state, and they expressed that they weren't looking for feedback yet and that's why they had pushed it as wip14:45
ildikovtonyb: yes, that is the patch14:45
ildikovClark[m]: it wasn't pushed as WIP, I rebased it on top of master on the GUI allowing conflicts. Which was probably a stupid choice, but I only realized after I ended up in a pickle with it.14:46
fungiit seems like they may have a culture of pushing their changes in wip state by default, and then turning off wip once they're comfortable with someone reviewing them14:46
fungiildikov: the original patch bruce pushed was pushed in a wip state, and that persisted onto your revision14:47
ildikovfungi: but it showed 'Active' in Gerrit14:47
ildikovWhere is that visible otherwise?14:48
ildikovfungi: I mean the Gerrit GUI had the 'Active' label on it when I last checked14:48
fungivery odd in that case. gerrit's change log doesn't show anyone setting it to wip14:48
ildikovWIP showed up after it ended up in a merge conflict14:49
fungidid you rebase locally and push?14:49
ildikovno, I rebased on the GUI, and checked the chekbox allowing conflicts14:50
fungimaybe this is a new effect of the rebase button in the webui14:50
tonybThat's what I'm thinking14:50
ildikovyeah, that's my thinking too14:50
fungioh, i had no idea gerrit started allowing you to rebase changes when they had merge conflicts14:50
fungii think that must be a new "feature"14:50
Clark[m]Note all wip changes are in merge conflict due to a quirk of gerrit's interpretation of mergable changes. Wip is not mergable so it shows as in merge conflict14:50
fungiin that case, i agree it probably auto-sets wip state on the revision it creates14:51
Clark[m]I'm not sure it was actually in merge  conflict14:51
ildikovtonyb: did you end up putting up a new change?14:51
fungiwell, in this case it rebased onto a nonexistent parent14:51
ildikovClark[m]: it was, it modified all the files that Bruce originally submitted with the merge conflict headers. I removed those already14:52
tonybildikov: that change doesn't contain a DCO. As the 900100 also doesn't have one due to the rebase/edit history14:52
ildikovtonyb: I think that's fine, I just thought y'all can fix the one that's stuck in WIP14:53
tonybildikov: Yup I pushed a new one.  I can also push onto 900100 to see what happens, but I suspect as I'm not the owner it will stay in WIP14:53
ildikovotherwise I would've cherry-picked it myself :)14:53
tonybAhh sorry I miscommunicated that :/14:54
fungiah, okay, the parent isn't nonexistent, gerrit's commit id search link is simply broken. 7878c22 is the penultimate change on the branch at the moment14:54
ildikovfungi: +114:54
fungiand just to be clear, bruce isn't around to un-set the wip state on 90010014:56
ildikovfungi: if he will attend the StarlingX TSC meeting then I can ask him to try to do that. Otherwise I can drop him an email and see. Is this a button on the GUI for him?14:57
Clark[m]fungi: commit sha search doesn't work for merge commits if the parent is a merge commit14:58
ildikovI'm just asking how he can unset it, in case I'll need to explain the process.14:58
fungiildikov: in the top-right corner next to the rebase button he'll see "mark as active" with an eye icon next to it14:58
ildikovBut we also have another version of the change up now14:58
Clark[m]Because typically there is no 1:1 mapping of a merge commit to a change id14:58
ildikovfungi: cool, thank you!14:59
fungiClark[m]: oh, right, i get it. gerrit won't show the commit if it's not the hash of a change14:59
*** blarnath is now known as d34dh0r5315:00
fungiildikov: also, i'd avoid "allow merge-conflicts" in the future, since the change won't be able to merge if it has merge conflicts, and you'll need to resolve those locally and re-push the revision anyway15:02
fungi(if it has actual merge conflicts)15:02
ildikovtonyb: fungi: ok, I did catch Bruce and he fixed the WIP15:16
fungiawesome. i confirm it says "active" now15:17
tonybGreat.  I abandoned the one I made15:17
ildikovfungi: this one was a very easy fix. It was also surprising to see how git dealt with it, as there were no files in the folders where the conflicts appeared at the end15:18
ildikovtonyb: great, thank you15:18
ildikovfungi: so I only chose that option as I did see what conflict there would be in the first place15:18
fungiildikov: yeah, i have a feeling "allow merge conflicts" with the rebase button may just always set the change into a wip state15:19
ildikov*did NOT see15:19
ildikovfungi: yeah, that was the learning curve on my end as I was not aware of the possibility itself of that happening15:19
fungiwhich seems like an odd behavior if so, since normal users wouldn't otherwise be able to set someone's change to wip15:19
ildikovnow that I know, I would not choose that option again :)15:19
tonybildikov: You, or another election person,  may want to update the acl file to allow toggleWipState, like
ajaiswal[m]tonyb:  i guess we need to update here 15:40
ajaiswal[m]field toggleWipState =15:41
fungiajaiswal[m]: yes. you can see an example at
fricklerfwiw I've been using the UI rebase-with-conflict a couple of times and resolved the conflicts in the UI editor afterwards. makes it easier if I don't already have the repo locally. so I wouldn't want us to disable that feature15:59
fungifrickler: good to know, and i keep forgetting they added an interactive editor to the ui. have you also observed the behavior described above, where the rebase button with "allow merge conflicts" set switches the change into wip state?16:03
fungiif that's really how it works, my main concern is that the feature effectively allows people to switch a change to wip without the ability to switch it back to ready16:03
fricklerfungi: yes, it switches to wip if there are actual merge conflicts, which kind of seems useful to me, but I do see the point about diverging permissions, maybe that could be considered a gerrit bug16:13
tonybclarkb, fungi: meetpad in ~35mins ?17:26
* tonyb may wander for coffee17:27
clarkbyup I made it back in time17:45
tonybNice 17:46
clarkbI am here:
opendevreviewJeremy Stanley proposed opendev/irc-meetings master: Update OpenStack TC meeting chair
opendevreviewMerged opendev/irc-meetings master: Update OpenStack TC meeting chair
*** JayF is now known as JasonF18:53
*** JasonF is now known as JayF18:53
fungipopping out for a quick dinner, but should be back around 21:00 utc or so19:56
opendevreviewClark Boylan proposed opendev/system-config master: Improve etherpad functional testing
clarkbtheres the testing thing I wanted to do yesterday before getting distracted by the shiny utf8 glyphs20:56
tonybclarkb: I have a couple of questions before I head out for beer21:18
clarkbgo for it21:18
tonybI notice that the RAX VMs all have '-int' names and addresses.  I assume I shoudl just include them with the zone update, and get the IP from the openstack CLI on the bridge node21:19
tonyb.... Or I guess I could get it directly from the new mirror02 with 'ip addr'21:19
clarkbtonyb: oh yes the rax mirrors get dns names for the intenral addresses because traffic is so much quicker there21:20
clarkbya ip addr works too21:20
clarkbits the 10 net address that only routes within rax and we have test nodes connect to that because throughput is much better21:20
tonybOkay.  got it21:20
tonybNext, I tried to log into the wiki server to look at it.  I think I may have made a mistake with the key choice as it seems to be rejected by the sshd on wiki.21:21
tonybis there a known workaround21:22
tonybNote I can get through the hostkey exchnage the server is rejecting my key itself21:22
clarkbtonyb: ya I think that is beacuse it isn't properly managed so one of us need to add you manually21:26
clarkbI'll defer to fungi on that as he has done most of the care and feeding for it21:26
tonybOoooooo of course that makes sense.  It isn't in the inventory so ... the user didn't get created.21:27
tonybNo rush at all.  I found what I was looking for another way.  Apparently we're running REL_1.28, and there are offical containers for it. So something to play with later.21:29
fungiokay, back now21:34
fungiyeah, no real config management on wiki.o.o, i'll copy your key to iy21:35
fungitonyb: see if you can authenticate now21:43
tonybYup.  Thank you.21:44
clarkbthanks for taking care of that21:50
clarkbI got a new phone today because old one will stop getting security updates this month. And now I'm raging at my bank for getting 2fa all wrong21:51
fungii even matched up the uid/gid21:51
clarkbanyone know why this may be happening:
clarkbmaybe because isn't running in a shell and I'm trying to subshell?21:59
clarkbor maybe its the -O- ?22:00
tonybYeah I think you need a bash -c in there.22:00
fungi-qO- should send all content from the request to stdout22:01
tonybI hit a similar issue when debugging the registry update22:01
fungibut yeah, i'm unclear on why that wouldn't make it into cmd.stdout22:02
opendevreviewClark Boylan proposed opendev/system-config master: Improve etherpad functional testing
clarkbif that works I'll have to think about formatting the commands a bit nicer22:02
fungiright, could be lack of a shell to handle the redirection, i guess22:02
clarkbalternatively I could do mroe of that in python but I watned to keep the command the same as our admin examples as good coverage for that too22:03
tonyblooks like you need whitespace after the -c22:42

Generated by 2.17.3 by Marius Gedminas - find it at!