Conversation

Notices

1. Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 29-Apr-2024 00:27:34 JST Haelwenn /элвэн/ :triskell:

   • pistolero

   @p

   I don't recall your age; I think you are younger than me, but if you are much younger than me, you might not remember that they have been doing clickbait headlines since there were no clicks, just physical paper.

   I'm from 1998 so somewhat on the edge I guess, like I know tabloids and celebrities magazines are a great example of offline buybait newspapers, but online newspapers where already a thing in my childhood.
   (And I guess they were long used to it due to the Minitel (french national BBS) being so widespread)

   ** (EXIT) no process: the process is not alive or there's no process currently associated with the given name, possibly because its application isn't started

   That's a weird one, I don't think I've hit it before. Maybe diff a file like lib/pleroma/application.ex (IIRC only one starting processes like this one but could be wrong) between current version and yours?

   • pistolero (p@fsebugoutzone.org)'s status on Monday, 29-Apr-2024 00:27:30 JST pistolero

     in reply to

     • pistolero
     @lanodan Guess that didn't work:
     
     14:09:42.450 request_id=F8p274v5jzTiCSIAAAEH [error] Internal server error: {:no
     proc, {:gen_statem, :call, [#PID<0.870.0>, {:perform, "/tmp/plug-1714/multipart-
     1714313379-840534496075925-7"}, 30000]}}
     
     14:09:42.457 [error] #PID<0.1497.0> running Pleroma.Web.Endpoint (connection #PI
     D<0.1496.0>, stream id 1) terminated
     Server: fsebugoutzone.org:80 (http)
     Request: POST /api/v1/media
     ** (exit) exited in: :gen_statem.call(#PID<0.870.0>, {:perform, "/tmp/plug-1714/multipart-1714313379-840534496075925-7"}, 30000)
     ** (EXIT) no process: the process is not alive or there's no process currently associated with the given name, possibly because its application isn't started
     
     I could really wish for a stack trace. At any rate, *some* uploads made it in, and according to the logs they got there via `PATCH /api/v1/accounts/update_credentials`, which accepts media but does something different with it or some AOP declaration is present in the regular uploader endpoints but not that one.
   •  (mint@ryona.agency)'s status on Monday, 29-Apr-2024 00:27:30 JST 

     in reply to

     • pistolero
     @p @lanodan A shot in the dark, but do you have exiftool installed? I reckon Pleroma throws an error upon launch if it can't find it, but considering your geriatric version, it might be running fine until it tries to process the upload.
   • pistolero (p@fsebugoutzone.org)'s status on Monday, 29-Apr-2024 00:27:31 JST pistolero

     in reply to
     @lanodan
     
     > like I know tabloids and celebrities magazines are a great example of offline buybait newspapers,
     
     I mean the "respectable" papers did that kind of thing, like all of them. "IS THERE LIFE ON MARS?!" in the headline, and "Well, we found some methane, but nothing indicating life." in the text. Michael Crichton had a very clever rant about this.
     
     The papers are still sloppy and irresponsible when they weren't outright deceptive, the main thing that changed is that they're summarizing things the journos saw on Twitter rather than things that they went outsite and looked at or possibly even investigated.
     
     > That's a weird one, I don't think I've hit it before. Maybe diff a file like lib/pleroma/application.ex (IIRC only one starting processes like this one but could be wrong) between current version and yours?
     
     AH! Okay, excellent, thanks. That is already helpful. I've read the diff a few times, and I did yank Prometheus but I don't know which process it isn't finding alive; I think Prometheus would crash the poor feller on other endpoints besides just media uploads. So there must be some difference between that and other stuff, or maybe the crash is occurring as part of an attempt to log a different failure elsewhere.
     
     I had an idea, but then realized I should probably send this before recompiling and restarting in case it's a disaster.
   •  (mint@ryona.agency)'s status on Monday, 29-Apr-2024 00:43:11 JST 

     in reply to

     • 
     • pistolero
     @p @lanodan Also see if you could simulate the upload pipeline in iex, similarly to how my MRF uploads media.
     https://gitgud.io/ryonagency/pleroma/-/blob/ryona-dev/lib/pleroma/web/activity_pub/mrf/inginsub_avatar_change_tracker.ex?ref_type=heads
   •  (mint@ryona.agency)'s status on Monday, 29-Apr-2024 01:18:57 JST 

     in reply to

     • 
     • pistolero
     • ​
     @lanodan @p Checked my own pleromer tree, avatar/banner/background uploads in Pleroma.Web.MastodonAPI.AccountController.update_credentials request go through a bunch of hoops before landing on Pleroma.Web.ActivityPub.ActivityPub.upload, the exact same function that's called directly by Pleroma.Web.MastodonAPI.MediaController.create. The first difference between Pleroma.User.put_upload and ActivityPub.upload I see is that the former calls Plug.Upload directly while the latter seems to be using some assigns from Majic.Plug. So, exiftool is a false lead since it's either called or not called on both type of media uploads.
     Updated banner on @mintplague also doesn't load (throws 502 after a bit of waiting), but that's the issue beyond Pleroma's scope as it accepted the upload fine.
   •  (mint@ryona.agency)'s status on Monday, 29-Apr-2024 01:25:26 JST 

     in reply to

     • 
     • pistolero
     • ​
     @lanodan @mintplague @p I still think that a missing library might be a culprit.
     https://git.pleroma.social/pleroma/elixir-libraries/majic/-/blob/develop/README.md?ref_type=heads
     >You must also have libmagic installed locally with headers, alongside common compilation tools (i.e. build-essential). These can be acquired by apt-get, yum, brew, etc.
   •  (mint@ryona.agency)'s status on Monday, 29-Apr-2024 01:41:24 JST 

     in reply to

     • 
     • pistolero
     • ​
     @lanodan @mintplague @p Exact same error. I'm 100% sure you just forgot libmagic.
     https://git.pleroma.social/pleroma/pleroma/-/merge_requests/3168
   •  (mint@ryona.agency)'s status on Monday, 29-Apr-2024 03:35:24 JST 

     in reply to

     • 
     • pistolero
     • ​
     @lanodan @mintplague @p >difference between Pleroma.User.put_upload and ActivityPub.upload
     *difference between Pleroma.User.put_upload and MediaController.create
   • pistolero (p@fsebugoutzone.org)'s status on Monday, 29-Apr-2024 20:13:24 JST pistolero

     in reply to

     • 
     @mint @lanodan I've got the EXIF-stripping off and have been testing with a text file, but to be certain, I tried this anyway. It didn't work, but it gave me the idea that, since I've been testing with a text file, maybe I should check libmagic:
     
     $ file deps/majic/priv/libmagic_port
     deps/majic/priv/libmagic_port: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 3.2.0, with debug_info, not stripped
     
     That shouldn't exist: I did deps.clean and rebuilt all of them, specifically to avoid this problem. So then `find deps -type f -executable | xargs file | grep x86-64`. Most of the executables in there are indeed aarch64, except two: deps/fast_html/c_src/lexbor/CMakeFiles/3.19.3/CMakeDetermineCompilerABI_C.bin and deps/fast_html/c_src/lexbor/CMakeFiles/3.19.3/CMakeDetermineCompilerABI_CXX.bin. I'm not sure what lexbor is, but I imagine that if it was even in use, that way more shit would have broken if we were actually using it anywhere.
     
     Anyway, check it out:
     geniusretard.jpg
      likes this.
   • pistolero (p@fsebugoutzone.org)'s status on Monday, 29-Apr-2024 20:13:24 JST pistolero

     in reply to

     • 
     • pistolero
     @mint @lanodan Oh, right, I didn't do the other thing yet. I gotta whitelist this domain in the media server.
     
     But also thank you. I didn't even think about the deps until you said this, because I did a clean/compile on them. (I had to edit some of them in place because of build issues from using an older version of Elixir but a newer version of openssl.)
      and † top dog :pedomustdie: like this.
   •  (mint@ryona.agency)'s status on Monday, 29-Apr-2024 20:39:26 JST 

     in reply to

     • pistolero
     @p @lanodan >Anyway, check it out:
     Celibate salmon.
     
     Looks like my hunch was true, though I still assumed it's a missing dependency instead of an incorrectly installed one. Never seen that happen myself, Larry Ellison's laptop that ran cum.salon was aarch64 as well, and Pleroma had no trouble running on it. Might that be a leftover from when you copied the installation to the new machine?
     † top dog :pedomustdie: likes this.
   • pistolero (p@fsebugoutzone.org)'s status on Monday, 29-Apr-2024 23:16:07 JST pistolero

     in reply to

     • 
     @mint @lanodan This is an extremely useful thing to do and I may have to refer to it at some point; thank you.
      likes this.
   • pistolero (p@fsebugoutzone.org)'s status on Monday, 29-Apr-2024 23:26:44 JST pistolero

     in reply to

     • 
     • ​
     @mint @lanodan @mintplague
     
     > So, exiftool is a false lead
     
     It was what got me to the answer, though. "Indirect lead" maybe.
     
     > also doesn't load
     
     Yeah, it won't.
     
     > (throws 502 after a bit of waiting)
     
     Because media.fse timed out trying to get to regular.fse.
      likes this.
   • pistolero (p@fsebugoutzone.org)'s status on Monday, 29-Apr-2024 23:27:25 JST pistolero

     in reply to

     • 
     • ​
     @mint @lanodan @mintplague Yeah, the issue was that it was not missing but built for the wrong architecture.
      likes this.
   • pistolero (p@fsebugoutzone.org)'s status on Tuesday, 30-Apr-2024 17:07:54 JST pistolero

     in reply to

     • 
     • ​
     @mint @lanodan @mintplague Almost; I didn't get any errors about missing symbols. Literally nothing except SSL errors (harmless hackney-generated warning) on either side of the stuff I pasted.
      and † top dog :pedomustdie: like this.
   • pistolero (p@fsebugoutzone.org)'s status on Tuesday, 30-Apr-2024 20:07:02 JST pistolero

     in reply to

     • 
     @mint @lanodan
     
     > Never seen that happen myself, Larry Ellison's laptop that ran cum.salon was aarch64 as well, and Pleroma had no trouble running on it.
     
     Is it not on Larry Ellison's laptop any more?
     
     Anyway, it's got no trouble, just there was some dependency hell finding libraries that worked with old Pleroma and new openssl and a version of Elixir that hadn't deprecated a bunch of shit (I do not know how Pleroma devs deal with this level of churn; a program that was written two years ago doesn't work in current Elixir because they continuously rename shit, not even functionality changes, just senseless fiddling without regard for maintenance or versioning or any of it), and apparently deps.clean doesn't clean all of the deps.
      likes this.
   •  (mint@ryona.agency)'s status on Tuesday, 30-Apr-2024 20:10:34 JST 

     in reply to

     • pistolero
     @p @lanodan >Is it not on Larry Ellison's laptop any more?
     Pernia's account got suspended for reasons unknown. It's been on the same chinkcentre box as agency for half a year at this point. Proxied to my frantech VPS exactly the same way as well.
     New Janny in Town likes this.