#social 2017-10-27
2017-10-27 UTC
timbl, rowan, dlehn and xmpp-social joined the channel
# 
puckipedia rhiaro: hahaha. same issue, after JSON-LD compacting
# puckipedia I wonder if this was an oversight
# puckipedia also yeah, really didn't realise lol
# puckipedia oh. I think I know.
# puckipedia rhiaro: totalItems is defined as http://www.w3.org/2001/XMLSchema#nonNegativeInteger
# puckipedia rhiaro: fixed it by defining type lol
# puckipedia I implemented the JSON-LD API myself, and remembered that the inverse mapping is also sensitive to type
# puckipedia yay :)
# puckipedia I need to develop a database migration tool - older objects in my DB still have a plain AS2 @context
# puckipedia my new context takes the Mastodon context and adds a few Kroeg extensions to it, like fixing up jwks/uploadMedia/likes/liked as those aren't defined in AS2. then it adds kroeg:settingsEndpoint and kroeg:blocks
# puckipedia but the @context in the DB overrides these extensions so now older objects have _:likes
# puckipedia egh. just figured out why published etc weren't set
# puckipedia if you guessed "forgot to declare type" then yes
# puckipedia slowly building a triple store
rowan joined the channel
rowan, rowan_ and timbl joined the channel
# erincandescent cwebber: I feel it's actually good for a test suite to often do the minimally correct thing
# erincandescent You should mix it up if possible :)
# puckipedia erincandescent: though,
{"type": "Like", "object": "[...]"} is probably a bit too minimal# erincandescent Hmm, could do with an actor :p
rowan_ joined the channel
# puckipedia yeah. my code decided "this is an activity. there's no actor. not allowed."
# puckipedia and this was basically cause for most of the tests failing. also something with access control, might be on Kroeg's end
# puckipedia (if an object has an actor, it won't get wrapped in a Create. but I did hard-code a list of definite-activities that just get refused without actor. for extensibility, of course :P)
# cwebber puckipedia: I assumed the server fills in the actor :)
# cwebber in c2s
# cwebber s2s is another matter!
# cwebber after all you probably throw out the actor if it's wrong anyhow in c2s
# puckipedia I actually just refuse
# puckipedia but, this is more to prime client developers to do proper "actor" adding
# cwebber :)
# puckipedia so if they decide to add custom Activities they will also pass `actor` and they won't get weird behaviour
# cwebber puckipedia: fair
# cwebber I'll add that in a few
# puckipedia yeah, no hurries, still working on my triple store :)
# puckipedia actually I think I just finished it
# cwebber wow!
# puckipedia I figured out how to store entities with no ID in the store. it's a really ugly hack
# puckipedia (I namespace all the triples into an TripleEntity object, while also having a separate Subject)
# puckipedia okay, I overlaid the triple store on top of the normal DB store
# puckipedia god. GETting the outbox is now up to 5 seconds
# cwebber oop
# puckipedia tbh, this is from a 'cold start'
# cwebber puckipedia: I assume you've added the appropriate indexes?
# puckipedia not yet lol :D
# cwebber well that probably affects things ;)
# puckipedia oh lol. I think it actually stored into the DB. but adding to the collection failed
# puckipedia woo! triple store has roundtrippyness
# puckipedia ... almost
# cwebber puckipedia: cool :)
# cwebber puckipedia: are you going to use triples or quads?
# puckipedia ... I really just designed quads, didn't I
# cwebber I'm a bit too fuzzy right now to tell :)
# cwebber puckipedia: btw I just added an actor field to every activity in the test suite
# puckipedia well, I basically have a 'graph' which is the main object and sub-objects inside a single Entity
# cwebber sounds like quads :)
# cwebber that's good, you want quads
# puckipedia okay so I developed quads. great.
# cwebber puckipedia: how are you extracting the "object shape" from it? are you going to use json-ld framing?
# puckipedia currently I can't recreate arbitrary document shapes yet
# cwebber json-ld framing probably already does what you want
# cwebber take a look and check at least
# puckipedia tbh I'd vote for signing every object separately
# cwebber > Submitted object, but no ActivityStreams object at response's Location
# puckipedia yeah that's the 403 issue
# cwebber oh
# cwebber I probably am not retrieving the objects, despite submitting them
# cwebber under the client's credentiasl
# cwebber easy fix
# puckipedia okay, I'm throwing the test suite against my quad-based store
# cwebber let's see
# puckipedia welp I got disconnected :P
# cwebber oh
# cwebber I was restarting it sorry
# cwebber since I was switching some of the requests to use the client's credentials
# cwebber back up now
# cwebber well that's a lot more Yes than before!
# puckipedia first 4: yes. froze at the response for "Up for some root beer floats?"
# cwebber well we're getting closer ;)
# puckipedia as in, no response after the GET request thing
# cwebber puckipedia: on your end or mine?
# puckipedia your end
# puckipedia retrying tests now
# puckipedia now it works(TM)
# cwebber nice :)
# cwebber oh yeah the "froze" was because I restarted
# cwebber mid-process
# cwebber heh!
# puckipedia "name": "An indecisive note"
# cwebber hehe
# puckipedia that one is still missing an actor
# cwebber oh ok
# cwebber there are some almost-jokes in the activities
# puckipedia and the following doesn't return the follow because it hasn't been Accepted (also froze again)
# cwebber puckipedia: should be fixed
# cwebber oh well re: the froze... another continuation bug :|
# cwebber ERROR: Attempt to suspend fiber within continuation barrier <- bane of my existence
# cwebber that could have been my fault this time for live hacking the code during the middle of it running
# cwebber puckipedia: could you try one more time?
# cwebber ERROR: Attempt to suspend fiber within continuation barrier
# cwebber argh
# puckipedia lol. it's still running though
# cwebber I'm lost as to why these continuation bugs are happening with your server but not mine
# cwebber obviously not your fault since it's a language issue
# puckipedia I respond with the object contents on 201
# cwebber I don't think that's related if you're suggesting that's why
# cwebber I do the same thing
# puckipedia hm
# cwebber the "root cause" of the bug is described here https://wingolog.org/archives/2010/02/26/guile-and-delimited-continuations
# cwebber what it really means
# cwebber is that the continuation is doing some clever things so I don't have to write callback style code because I hate that
# puckipedia do you know where it happens
# puckipedia bbiab, food
# cwebber which is all good and well but it involves the language doing something where, above a "prompt" layer, it suspends the code so the scheduler can resume it when it's ready
# cwebber the problem is it doesn't work if you "mix in" some C stack
# cwebber but afaict I'm doing all scheme, and at least it shouldn't be different
# cwebber puckipedia: sadly not quite, it doesn't "pinpoint" where the issue is happening... which is why this is the hardest damn thing to debug
# cwebber I may have an idea on how to find it
# cwebber maybe, maybe I can actually walk the stack and keep checking for the point at which the error is introduced
# cwebber this might be possible.
# cwebber but
# cwebber OH uh
# cwebber I wonder if my "fix" for the async https stuff is partly causing the issues here
# cwebber of course.. I'm doing call-with-new-thread... that's *almost certainly* the cause.
# cwebber I think I have an idea on how to fix it that's less hacky than my current solution.
# cwebber Oh! This may be easier than I thought.
# cwebber saying it here so I don't forget:
# cwebber have each <case-worker> (test suite runner) run in its own independent thread instead of breaking out one to try to get around the https issue.
# cwebber that should be good enough.
# cwebber I need to pack up. I'm at a wedding this weekend. I'll work on closing the remaining AP issues while on the trip though.
# cwebber fixing the continuation bug may have to wait until monday however... we'll see how much time I get.
# cwebber puckipedia: thank you again for all your help.
# cwebber you have really done a great job.
# cwebber also thanks rhiaro for all the help recently on closing issues!
# xmpp-social [ajordan] Have fun cwebber! \o/
# cwebber thank you ajordan :)
# cwebber I'm restarting the test suite on test.ap.rocks (ie, shutting it down from my laptop and putting it back up on the server proper, for obvious reasons)
# cwebber test.ap.rocks now running on the server proper again
rowan, rowan_, alanz and timbl joined the channel
alanz_ joined the channel
# rowan_ cwebber: i'm confused about how to use test.activitypub.rocks. is it just a series of questions, or is it supposed to do more?
# puckipedia I should be properly caching ID values now
# puckipedia bleh. need to basically rewrite the entirety of the collection utils. maybe replace it with a list of quads? :P
# puckipedia also I can garbage collect the contents of entities while keeping metadata intact, this could be cool
# xmpp-social [ajordan] Why would you need metadata but not data? Except for Tombstones?
# puckipedia I store all remote entities as well, I could possibly make a mechanism to purge them okay I have no reason to do so
# xmpp-social [ajordan] Lol
# xmpp-social [ajordan] Gotcha
timbl joined the channel