ByteCode.NewsSubmit
Use quotes for an exact word.

Subscribe

RSSPodcast
Log in

Chat Logs

  1. phalethclanker dropped a ton of sh script, for some reason it still keeps some test sh scripts https://repopack.com/project/repopack/tasks/782
  2. phalethactually some of that is bash
  3. phalethalso clanker should have developed rps against golang 1.27, not golang 1.24 which is obsolete
  4. blueyes, also hi
  5. phalethhi blue
  6. bluehi
  7. bluerepopack.com is down, I'm doing some maintenance, should be back up soon
  8. blueshould be back
  9. bluemost of the scripts are gone
  10. blue[blue@vmd170320 ~]$ ./rps
  11. bluerps: usage: rps <bootstrap|status|sync|deploy|rollbac
  12. blueyou can build rps and copy it; but you'd need a ~/.ssh/repopack_deploy_ed25519 to deploy
  13. bluedreamreal: clankers are overengineers of the worst possible kind
  14. blueI feel like phaleth's input is valuable. we might be soon needing programming languages that work well for clankers
  15. bluephaleth: https://repopack.com/project/repopack/tasks/786
  16. phalethdidn't know task can be renamed
  17. phalethcool, renamed a closed task
  18. blueyes
  19. bluephaleth: everything is so much faster and nice with go rps now
  20. bluethat mvp thing was a terrible calnker hallucination
  21. phalethyeah, I bet
  22. blueI think clanker should be 100% forbidden from writing a single code of bash
  23. bluemaybe only go
  24. phalethanyway, it seems like rps depends on a set of CLI tools
  25. bluewhat do you mean with CLI tools?
  26. phalethyeah, I guess maybe you could put that into AGENTS.md or whatever it's called
  27. blueya
  28. phalethit looks like the rps code is calling into a lot of what the shell scripts used to do
  29. phalethlike tar and whatnot, instead of using golang standard lib
  30. phalethone thing it does correct is that it uses the podman REST API
  31. phalethit's too POSIX dependent it seems
  32. bluePOSIX is the way to go no
  33. blue?
  34. phalethnot when standard lib already has a solution
  35. phaleththe best approach is of course always stay within the process
  36. bluedeclankered https://repopack.com/project/repopack/tasks/792
  37. phalethof course there are OS kernel calls, but those should also be minimized by teh runtime
  38. blueya
  39. phalethand anyway, writing to disk on a server if not really necessary is the dumbest thing a programmer can allow to happen
  40. phalethif there is a streaming API then it's a no brainer
  41. bluewe need a git streaming API :P
  42. phalethyeah, clanker has some really tough task on repopack
  43. bluehm?
  44. phalethwell, there is a lib for storing git data into postgres
  45. phalethforgot what it was called
  46. blueya
  47. phalethhopefully that one uses UUIDv7s everywhere by default
  48. bluethat's another can of worms we'll have to handle, that uuidv7 migation
  49. bluebut at least we don't have those terribly bash scripts anymore, what a nightmare
  50. phalethyeah, should do it sooner rather than late
  51. phalethreplacing 50k lines of shell script with 4k lines of golang is not bad
  52. blueyup
  53. phalethblue: sshd container is created but not started
  54. phalethcreated 32 Minutes ago
  55. bluethanks phaleth, looking into it
  56. phalethno problem
  57. bluephaleth: sshd is fixed
  58. phalethok, cool
  59. dreamrealblue: true but because humans are overengineers too
  60. blueya
  61. bluedreamreal: could you look at what I asked on #primate please?
  62. bluephaleth: I wanna start working on the container registry. we need a short domain. any suggestions?
  63. bluerpcr.io ?
  64. phalethcan start with cr.repopack.com and buy another domain later
  65. blueok. I don't mind buying the domain. I believe the shortness is that image names are short
  66. phalethyeah, but you are not going to build a docker hub like website just yet right?
  67. bluewell, we're going to do a proper container registry, no? like gcr.io
  68. bluerpcr.io/PROJECT/IMAGE
  69. phalethI mean once you have somthing up and running and something depends on it why not
  70. phalethbut till then
  71. phalethwho knows how all these oci-image registries work anyway
  72. bluewell, if we're doing one, we should do it seriously. PROJECT should be scoped privately if the project is private, and so on
  73. blueyou suggested having a CR so that we don't need to depend on local caching of images
  74. phalethyeah, but can the registry be distributed to multiple VPSes?
  75. bluedistributed in what sense? it will be constructed such that it's publicly available. VPSes running user code will need to get a token that allows them to pull from the registry, scoped to the given project
  76. bluepublic projects could offer public images
  77. phalethsay there are two VPSes in different geo zones, rpcr.io points to both, now a user in zone A uploads new image or image version, how will the registry in zone B obtain that image?
  78. phalethif repopack controls that process then it's just db record being copied and also a tar.gz file being copied and imported to podman on target VPS
  79. phalethanyway, for now rebuilding images is fine, it's not being done too often and it's rather quick considering that the only optimization that happened is flog
  80. blueyes, it's my intention that it's part of repopack's database
  81. phalethI'd highly recommend to focus on all those already existing tasks first
  82. phalethmaybe even squeeze primate a bit more for some optimizations
  83. bluedo you have concrete suggestions?
  84. phalethcode splitting, ditching svelte, dropping the bloat off of auth pages, torturing clanker into doing more with modern css than with js
  85. phalethI mean the app js bundle dropped from 1.81 MB to 1.13 MB with primate 0.43, but the site is far from feature complete and it will get back there eventually
  86. blueok, can you create a task with all those things? I can work them off one by one
  87. phalethsure, I can try
  88. bluethanks
  89. phalethanyway, do you think existing tasks are not worth it?
  90. phalethhttps://repopack.com/project/repopack/tasks/679
  91. bluethere are many existing tasks
  92. blueI'm working on many fronts at once
  93. phalethyeah, heh, I know
  94. phalethit's insane
  95. blueyup
  96. phalethhttps://repopack.com/project/repopack/tasks/794
  97. blueyes?
  98. bluedreamreal was 100% against this
  99. bluethat's why it's generally in the icebox now until someone raises his voice in support
  100. phalethnot sure what you mean
  101. phalethagainst code splitting?
  102. phalethprimate website is code splitted and it's working fine
  103. blueI mean 794
  104. blueoh wait
  105. blueI was looking at 784
  106. bluesorry
  107. blueyeah, I'm not sure we have such granular control over splitting. splitting occurs via groups, but we can try
  108. phalethyeah, I think just routes is good enough
  109. bluek
  110. phalethno clue what's 784 about, feel free to close I guess, up to you
  111. blueya, will probably close
  112. phalethI think the less features and fly out menus and UI crap are there the better
  113. phalethfor example the notification thing is very annoying
  114. phalethand it does not work well anyway and testing of that is pain in the butt
  115. bluewhat notification thing?
  116. blueyou mean desktop notification or bells?
  117. phaleththe bell thing in top left area
  118. bluehow else can I know that I need to do something somewhere?
  119. blueif it's not working well we can fix it
  120. phalethyou open it when you want to know
  121. phalethalso it notifies about bullshit, for example did it notify you about this? https://repopack.com/project/flog/apps/flog
  122. phalethI don't think so
  123. blueit didn't, show it? anyway you can opt out of notifs to my knowledge
  124. blueshould it*
  125. phalethoh, so let me turn that off then
  126. bluebut if something needs to be improved, please create a task instead of completely turning it off
  127. bluelike if you think something needs to be informed or not informed
  128. phaleththere are already so many tasks and it's not even the most important stuff :)
  129. phalethyou can't even look through all those tasks yourself, need a clanker to do that
  130. blueit's slowly improving, it's just a big scope, we've already got so much done, don't worry
  131. phalethyeah, the codebase is huge already
  132. blueit's actually not installed I believe, but was in my HOME
  133. phalethI guess there was a trust me bro installer doing that
  134. phalethand that trust me bro installer should have uninstall option
  135. phalethwell, if it's not from pacman than it does not matter
  136. blueit's not
  137. phalethso much RAM free after all those cleanups this week
  138. phalethabsolute waste, repopack app can grow ten times bigger
  139. blueya well, now just need to get rid of granafa and all the java dumpster fire
  140. phalethjava is gone
  141. phalethgrafana is not going anywhere, it'll only be moving
  142. blueI don't even know what grafana does and why it needs so much memory
  143. phalethwe only have one instance of prometheus so far, will prolly have a couple more
  144. phalethgrafana is a beast system, good thing it's written in golang
  145. phalethit's also amazing once you understand it at least a bit
  146. * phaleth still needs to expand those two HAProxy backend dashboards
  147. bluewhat is prometheus?
  148. phalethtime series data provider
  149. phalethalso written in golang
  150. bluewhy do we need all this complexity
  151. phaleththey work very well and actually not difficult to learn
  152. phalethlook into vaultwarden for grafana URLs and take a look at the dashboards yourself
  153. phalethall we are after is total automation and dashboards and then more automation to be able to quickly recover crap we run
  154. blueya
  155. phalethonly via well crafted dashboards we will be able to achieve the goal
  156. blueyeah let's just not fall in love with numbers like americans do
  157. bluenumbers are meaningful, but they're not everything
  158. blueI'll look into grafana
  159. phalethgrafana and prometheus do very well on one important metric and that is their RAM usage
  160. phalethcurrently there is not so much data in prometheus right now, cause I still didn't fully figure out backups, but grafana has no problem displaying one month of data from prometheus
  161. bluek