Monday, 2026-03-30

*** ralonsoh_ is now known as ralonsoh11:36
opendevreviewThales Elero Cervi proposed openstack/project-config master: Add repo app-prometheus to StarlingX  https://review.opendev.org/c/openstack/project-config/+/98249513:11
opendevreviewThales Elero Cervi proposed openstack/project-config master: Add repo app-prometheus to StarlingX  https://review.opendev.org/c/openstack/project-config/+/98249513:22
*** ykarel_ is now known as ykarel14:02
*** ykarel is now known as ykarel_14:14
*** ykarel_ is now known as ykarel14:15
opendevreviewMerged openstack/project-config master: Add repo app-prometheus to StarlingX  https://review.opendev.org/c/openstack/project-config/+/98249519:12
sean-k-mooneyclarkb: by the way remember the issue i raised with gerrit not displaying all comment when navagating between reviews. that is still happening today21:28
sean-k-mooneydo we think gerrit 3.12 or the cache backend changes will fix it21:29
sean-k-mooneyas far as i coudl tell it was an issue in the gerrit react applcation itself 21:29
clarkbok I remember spending a lot of time trying to reproduce it and I thought we decided the problem had to do with gerrit autohiding things across patchsets21:29
sean-k-mooneyyes21:29
sean-k-mooneybut i was also seeing it on the latest revision of a patch today21:29
clarkblike if you're looking at the wrong patchset version for teh comment. SO I think it is "intentional" (whether correct is another story) so I doubt that the cache change would help21:30
clarkbit is possible that they have fixed it as part of the UI updates though21:30
sean-k-mooneyin this case i was looking a the latest revsion of a patch where i had draft comment pending submission21:30
sean-k-mooneyand my draft comments were not showing21:30
clarkbwas the draft comment for an older patchset?21:30
sean-k-mooneyuntil i hit F521:30
sean-k-mooneyno 21:31
sean-k-mooneyi may have got do an older revsion at some point and gon back to the latest but when i did next and prevs it didnt fix it21:31
sean-k-mooneyas in  the app21:31
clarkbok then ya I doubt the cache backend update will affect that. I suspect it has more to do with how the web ui is deciding what data it needs/wants and whether or not it can proceed without that21:31
clarkbso it is possible newer Gerrit will help just not due to the cache change21:32
sean-k-mooneyso the reason i noticed it is i had the same review open in two diffent browser windows since i was review diffent fiels in both (test and production code i think)21:32
sean-k-mooneyand the comen i saw in oen browser were not lining up to the other21:32
clarkbunfortunately, I don't think I've ever observed the issue on my end. With all of the weirdness I saw being explained by looking at the "wrong" patchset21:34
clarkbif you see it again maybe check the network log for the page or the js console to see if there are any errors notes?21:34
sean-k-mooneylast time i looke there were none21:34
sean-k-mooneyso i think the issue is a related to how they redwa the screen when you change patches using next/previous or the keyboard shortcurt21:35
sean-k-mooneyi dont think its doing a full page refresh21:35
sean-k-mooneyi think they are doing ajax queries for the data and re rendering21:35
sean-k-mooneyin that flow it does not get all the comments21:36
sean-k-mooneyin a full page refrsh it does21:36
clarkbya that could explain it. Could also explain why I don't see the problem as I don't tend to change patchsets at the top level just compare within the diffs21:36
sean-k-mooneyi think i looked at that in the network tab in the past21:42
sean-k-mooneyso ya if you change files it just calls the diff endpoint 21:42
sean-k-mooneyhttps://review.opendev.org/changes/openstack%2Fwatcher-dashboard~978258/revisions/5/files/releasenotes%2Fsource%2Fconf.py/diff?intraline&whitespace=IGNORE_NONE21:42
sean-k-mooneywhich is just the json for that file21:42
sean-k-mooneybased on somehting (presumabel the content of the diff responce it sometall does a secod or subsequet network requet for content and fix:preview 21:45
clarkbI think that change details with the comments option is how you fetch the comments21:53
clarkbbut I'm not positive that that is how the web ui does it21:53
sean-k-mooneywell looking at the raw web responce in the borswer i actully cant see the network request for them at all21:54
sean-k-mooneyeven when it dispalying21:54
sean-k-mooneyso the are doing somethign tricky in the background21:54
sean-k-mooneyi can see the XHR ajax request for each page but its on tincldued in the respocnes21:55
clarkbit is the changes detail request I found it21:57
clarkbgive me a second to extract it out of firefox and I can share21:57
clarkbhttps://review.opendev.org/changes/openstack%2Fwatcher-dashboard~978258/detail?O=somevalue21:58
clarkbI don't know if the somevalue is user specific and should be kept secret so I edited it21:58
sean-k-mooney ack21:58
sean-k-mooneyso im not seeing it call that for what its worth when i navaget between paages21:59
clarkbso maybe it is caching the value in the browser with data at time A. Then you work in browser #2 and make new changes at time B where B>A. If the first tab isn't refreshed to reload that content then you don't see the new state22:00
sean-k-mooneyits loading it on the first refesh22:01
sean-k-mooneyand then its not querying it again22:01
sean-k-mooneyif you change version of navigate to a new file22:02
sean-k-mooneyit does not qurey that again22:02
clarkbright its using all the existing state and not requerying22:02
sean-k-mooneyyep which is invild22:02
clarkbI wonder how much of that is intentional vs oversight on their part. I knwo in the past they had the little yellow pop up thing telling you there is new conent and you can manually refresh22:02
sean-k-mooneyits not how it used to work in the past22:02
sean-k-mooneyas in when it was still a gwt applciations22:03
clarkbright but its been years since then. I wonder if the yellow helper box was intented to mitigate this and now that isn't working as expected for some reason either?22:03
sean-k-mooneythey have done some major rewrite but even when you click up to go back to the main view22:03
sean-k-mooneyit does not reload it 22:03
sean-k-mooneyso it very possible to leave review comemtn wihtout seeing feedback form others22:04
sean-k-mooneywhat yellow box are you refering too by the way?22:04
sean-k-mooneyi dont think i have ever seen that22:04
clarkbit is like a pop up thing that would say "change has been updated click here to refresh"22:04
clarkbit used to do that and now I wonder if this behavior is why22:04
clarkbI also wonder if that stopped working22:05
sean-k-mooneyi have seen that on the on the main page but never on any of the diff views22:05
clarkbyes on th top level change page not the diffs22:05
sean-k-mooneybut if they can send that notificaion they can requry the content so i would still conidre it a bug 22:05
clarkbbut I haven't seen that do it in a while22:05
sean-k-mooneythe only popup i have seen in gerrit recently is the one they do when you need to login again22:06
clarkband yes if they know the data is stale then they could query it. I know that Google in particular is very sensitive to UI responsiveness and wonder if this all goes back to them optimizing those metrics by not fetching data too often22:06
sean-k-mooneywell for me this is a ui resposivness fail22:06
clarkbin 3.13 or 3.14 they stopped making updates to the UI synchronous with backend actions so I'm sure we'll find even more weird cases22:06
sean-k-mooneybecuas it requrie a full applcation reset and often getting an email notficaton and going back to review comemtn that were there but you could not see22:07
clarkblike when you post comments it will immediately report success while the backend is workign on indexing them so no one will actually see them yet. Currently one reason taht posting comments or abandoning changes etc can be slow is it waits for the backend to report success before indicating success to you22:07
sean-k-mooneyright that sound like a terible idea22:08
clarkbapplication reset in this case is a refresh of the page22:08
sean-k-mooneythere have been cases in teh past where one person has +w and anohter has -1'd22:08
sean-k-mooneyso any non tansactionality is a problem22:08
clarkbI mean that can happen anyway22:08
clarkbI agree that this behavior is less than ideal. But Gerrit can't solve disagreements between reviewers generally22:09
sean-k-mooneyit can but ligitbally if i cant see your coment when im navagting between file or on the main view when im clickign reply22:09
sean-k-mooneythat a problem22:09
sean-k-mooneyrgiht but if its activly hiding the comemnt that a problem22:09
sean-k-mooneyas in if  i only use the naviationin the gerrit applciation22:10
sean-k-mooneywhewn i click reploy i cant be sure i have seen feeback for other that was there before i naviaged back to the main page22:10
sean-k-mooneyin the past it was alwasy visabel and you could confirm you didnt miss anything before leaving your review22:10
clarkbdoing some testing I'm able to reproduce what you are saying (which si good because we should be able to take that to gerrit if so) but only if I'm navigating within the change. As soon as I navigate back up to the project listing then back to the change it loads the data22:11
clarkbstill doesn't help if two people are reviewing the same change concurrently and are in that change view the entire time22:11
sean-k-mooneyrgiht so i would expect most pepoel review by navigating the chage diff then going back up to the main review without ever refressing the page22:12
clarkbyup or even just viewing the diffs on the main page22:13
clarkbloading the change in a new tag also seems to update things22:13
clarkbso its not a shared state across all tabs but seems to be tab specific?22:14
sean-k-mooneyyep so gerrit is expected to be a synconous transactional view of the db. any eventual consticy is stictly a bug in my view22:14
sean-k-mooneybut at the minim i think the navaction up to the main view should do a full refersh for other comments22:14
sean-k-mooneyclarkb: each tab has its own instance of the applcation running client side22:15
clarkbwhen you say main view you mean the top level chagne page right?22:15
sean-k-mooneyyes22:15
sean-k-mooneyi.e. https://review.opendev.org/c/openstack/watcher-dashboard/+/97825822:15
clarkbgot it. Just wanted to clarify because going up to the top level project page then back to the change does reload the info22:16
sean-k-mooneyhow did you go up did you click on the rep/branch linke22:16
sean-k-mooneyor one of the nave bar drop downs22:16
clarkbyes I clicked on the opendev/sandbox link in the top right of the change page next to Repo | Bnrach22:17
clarkbthen I clicked on the change from that link to go abck to the change and that updated with the new comments22:17
sean-k-mooneyya so that is hitting the detail enpoint22:18
sean-k-mooneymy guess is they have a react compont fo rthe review/diff view22:18
sean-k-mooneyand when we are naviagting within a view we are navagatin gin that compoent22:18
sean-k-mooneybut when we go up to the change view and back in its reloading it22:19
sean-k-mooneythere is a <gr-app id="pg-app"></gr-app> custom element22:19
sean-k-mooneyhum ok no its mroe then that22:20
clarkbit isn't react its something else, but yes probably operates similarly22:21
sean-k-mooneyit used to be gwt22:21
clarkbyes now it is "polygerrit"22:21
clarkbwhich is some custom fork of some other tool22:21
sean-k-mooneythe google web toolkit . it might be dart?22:21
sean-k-mooneybut ya22:21
sean-k-mooneyoh its using polymer22:23
clarkbI've left a tab open hoping that the stale tab will eventually notice it is stale but it hasn't22:23
sean-k-mooneyhttps://polymer-library.polymer-project.org/3.0/docs/devguide/feature-overview22:23
clarkbso yes I think that at a minimum here the bug is that gerrit isn't noticing and asking you to refresh. Ideally it would just load the new data when it notices it is missing in the first place22:23
clarkb3.14 is the version that makes web ui actions less synchronous with the backend (not 3.13)22:24
clarkband 3.14 hasn't released yet but will soon22:24
sean-k-mooneywhen i was workign at my university i ported a buch of freefrom javscript enrichmetn to a php sigte to polermer.js/dart22:24
sean-k-mooneypolymer was the main way to make webcomente just work across all browsers back in 201222:25
sean-k-mooneyim sort of happy it still exists22:25
sean-k-mooneyless so that it needs to exist as it really only exisit to make things work on older borwser that dont supprot modern web standards but im glad that the project is still maintained22:26
fungiit's relatively new in terms of gerrit webui implementations. this is, like, the third webui they've had now (second complete rewrite)?22:26
fungii suppose when they started working on the polygerrit interface, polymer wasn't all that old yet22:27
clarkbyes gerrit 3.0 was the first version with it iirc22:27
clarkbso its about 7 years old?22:28
sean-k-mooneywell polemer is actuly prtty nice as a way to od web-compoent in genreal 22:28
clarkbI need to pop out now to take advantage of a dry afternoon, but I think this is something we can put together into a bug for upstream22:28
sean-k-mooneythe browse api sfor this exsit but are not the nicest thing to work with normally22:28
clarkbbasically with gerrit 3.11.9 if you update a change in browser A and browser B already has the change loaded then nothing loads the new data in browser B or indicates that the data viewed is stale until you navigate away from the change and back to it again22:29
sean-k-mooneyyep ^22:29
clarkbI can file that. I'm also happy if someone else does it. I'll get to it tomorrow22:30
sean-k-mooneyyou might want to also clarify exactly what you mean by naviate away as it requrie mroe the next/up/prev and requires clicking oen of the other links in the application 22:31
sean-k-mooneyor like manually havatating a bookmark or something 22:31
sean-k-mooneyby the way im pretty sure the pollygerrit ui is using https://developer.mozilla.org/en-US/docs/Web/API/History_API22:33
sean-k-mooneyinternally to fake the history as if you had navigated22:33
sean-k-mooneythat why the url and history look like your naviagte but your actully not22:34
sean-k-mooneyits one of my favorte feature in htmx https://htmx.org/attributes/hx-push-url/ but it make the concept of "navigating" kind fo fuzzy22:34
sean-k-mooneyspecificly i woudl guess they are using https://developer.mozilla.org/en-US/docs/Web/API/History/pushState to enulate full page navagation without actully doing it22:36
clarkbyes, in this case the url needs to change from prefix/number to something else22:38
sean-k-mooney"this.prevPageContext=i,this.currentPath=i.path,this.dispatch(i,n),t&&!i.preventPush&&i.pushState()" 22:40
sean-k-mooneyso inlien jsvascript is not very readable but ya they are using that in the even handlers22:41
sean-k-mooneyah you can pretty pitn the source https://paste.opendev.org/show/b2FxgvkZY7gj68MLgDPZ/22:42
sean-k-mooneybefore "pretty" printing the file is ~25000 lines of js and embded html templsate after its ~113,000 lines of admittly much more readable code22:44
sean-k-mooneybut still not "pretty"22:44
sean-k-mooneyah22:47
sean-k-mooney pushState() {22:47
sean-k-mooney    window.history.pushState(this.state, this.title, this.canonicalPath)22:47
sean-k-mooney  }22:47
sean-k-mooneythat exactly what i was expecting to see22:47
sean-k-mooneyok ill stop distracting you. night all o/22:48

Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!