Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • 15:02
    raboof closed #28294
  • 15:01

    raboof on master

    actor: allow seamless access to… (compare)

  • 15:00
    patriknw milestoned #28274
  • 14:59
    patriknw closed #28274
  • 14:59

    patriknw on master

    Avoid dead letter for rebalance… Merge pull request #28274 from … (compare)

  • 14:52
    akka-ci commented #28318
  • 14:52
    akka-ci unlabeled #28318
  • 14:52
    akka-ci labeled #28318
  • 14:49
    patriknw milestoned #28303
  • 14:48

    patriknw on master

    Update sbt-paradox-akka to 0.28 Merge pull request #28303 from … (compare)

  • 14:48
    patriknw closed #28303
  • 14:28
    raboof commented #28308
  • 14:26
    akka-ci unlabeled #28128
  • 14:26
    akka-ci labeled #28128
  • 14:23
    raboof commented #28128
  • 14:23
    raboof edited #28128
  • 14:23
    raboof synchronize #28128
  • 14:23

    raboof on documentAsSubscriber

    Update sbt-paradox-akka To fix… (compare)

  • 14:22
    akka-ci commented #28315
  • 14:22
    akka-ci unlabeled #28315
Konrad `ktoso` Malawski
@ktoso
what a week :-) so much progress
Roland Kuhn
@rkuhn
well, the solution to that one builds heavily on SubSource and LocalSink :-)
(and subFusing, of course)
Konrad `ktoso` Malawski
@ktoso
:+1:
drewhk
@drewhk
@rkuhn we should make it possible for the users to access an unsealed HttpBlueprint if possible
i.e. if you want to fuse your handler into it
Roland Kuhn
@rkuhn
yes, definitely
drewhk
@drewhk
not for the masses, but for benchmarks it can be nice ;)
wheree is my taaaaaail soooource
ktoso @ktoso wohoo, new akka tshirt arrived!
drewhk
@drewhk
HttpServerSpec passes, yay
Konrad `ktoso` Malawski
@ktoso
nice :clap:
drewhk
@drewhk
let's see what fails still
I think the stance that "you shall consume your entity" is no longer acceptable. If we violate these regularly in our tests, then this is not something we can expect from users.
still 23 failures to go
drewhk
@drewhk
From now, I declare merge moratory on Akka Http
I am getting close to the finish with this and I don't want anyone to produce me new hours of debugging
Konrad `ktoso` Malawski
@ktoso
ok, sure :+1:
drewhk
@drewhk
whew, almost done!
we fuse!!
Matheus Lima
@matheuslima
:clap: :clap: :clap:
Konrad `ktoso` Malawski
@ktoso
wohoo! :shipit:
drewhk
@drewhk
ok almost, one deadlock is still there
dlock.png
above is the map of it. Seems like a cycle deadlock involving ResponseParsingMerge
btw, we should heavily optimize these blueprints, like collapsing maps into each other
to reduce memory usage
I guess this is another case of cycles in need of a prefetch
let's try
Roland Kuhn
@rkuhn
I think this one needs a buffer (if it works like the client-side one)
drewhk
@drewhk
this is the client side one
did you found this one too?
Roland Kuhn
@rkuhn
ah, yes: there should be a buffer(2) in front of the TerminationMerge
drewhk
@drewhk
why 2?
Roland Kuhn
@rkuhn
I am not certain, though: this is just from looking at the ASCII art diagram
drewhk
@drewhk
it can go in front of the parsingmerge too
will try both
love these deadlock diagrams
Roland Kuhn
@rkuhn
yes, they’re awesome :-)
Konrad `ktoso` Malawski
@ktoso
:+1: yeah, would be cool if we can provide them for people to use as well
Matheus Lima
@matheuslima
this deadlocks diagrams are generated using graphviz?
drewhk
@drewhk
dlock.png
this is when I add the buffer in front of responseparsing
still deadlocked, but now I see the doubly-linked cycle in this stuff. geez
Roland Kuhn
@rkuhn
yes; when I looked at the diagram I figured it should go on the other one :-)
(don’t ask me why, intuition)
@drewhk is your branch pushed as a PR?
drewhk
@drewhk
but terminationmerge is now not on a cycle