add an offline precompile workload #1195

IanButterworth · 2024-11-20T03:45:42Z

Fixes #1194
Or at least makes decent progress.

Before

julia> using HTTP
   1732.9 ms  ✓ HTTP (precompile time)

julia> @time HTTP.get("http://google.com");
  4.109167 seconds (13.00 M allocations: 658.440 MiB, 2.90% gc time, 96.96% compilation time: 2% of which was recompilation)

julia> @time HTTP.get("http://google.com");
  0.155600 seconds (11.27 k allocations: 746.781 KiB, 12.46% compilation time)

julia> @time HTTP.get("http://google.com");
  0.122175 seconds (574 allocations: 196.016 KiB)

This PR

julia> using HTTP
   6780.5 ms  ✓ HTTP (precompile time)

julia> @time HTTP.get("http://google.com");
  0.678762 seconds (1.17 M allocations: 64.638 MiB, 1.65% gc time, 77.77% compilation time: 86% of which was recompilation)

julia> @time HTTP.get("http://google.com");
  0.174314 seconds (11.28 k allocations: 746.500 KiB, 10.45% compilation time)

julia> @time HTTP.get("http://google.com");
  0.137826 seconds (596 allocations: 196.641 KiB)

All the big precompiles are baked in and just one big recompile left.

#=  339.1 ms =# precompile(Tuple{HTTP.var"##request#32", Nothing, Array{UInt8, 1}, Nothing, Base.Pairs{Symbol, Union{}, Tuple{}, NamedTuple{(), Tuple{}}}, typeof(HTTP.request), HTTP.MessageRequest.var"#makerequest#messagelayer##0"{HTTP.MessageRequest.var"#makerequest#1#messagelayer##1"{HTTP.RedirectRequest.var"#redirects#redirectlayer##0"{HTTP.RedirectRequest.var"#redirects#1#redirectlayer##1"{HTTP.HeadersRequest.var"#defaultheaders#headerslayer##0"{HTTP.HeadersRequest.var"#defaultheaders#1#headerslayer##1"{HTTP.CookieRequest.var"#managecookies#cookielayer##0"{HTTP.CookieRequest.var"#managecookies#1#cookielayer##1"{HTTP.RetryRequest.var"#manageretries#retrylayer##0"{HTTP.RetryRequest.var"#manageretries#1#retrylayer##1"{HTTP.ConnectionRequest.var"#connections#connectionlayer##0"{HTTP.ConnectionRequest.var"#connections#1#connectionlayer##1"{HTTP.TimeoutRequest.var"#timeouts#timeoutlayer##0"{HTTP.TimeoutRequest.var"#timeouts#1#timeoutlayer##1"{HTTP.ExceptionRequest.var"#exceptions#exceptionlayer##0"{HTTP.ExceptionRequest.var"#exceptions#1#exceptionlayer##1"{typeof(HTTP.StreamRequest.streamlayer)}}}}}}}}}}}}}}}}, String, String, Nothing, Array{UInt8, 1}, Nothing}) # recompile

Though there is an issue open about recompiles being incorrectly marked as such JuliaLang/julia#56155

Importantly, are these __init__()'s safe to call during precompilation?

IanButterworth · 2024-11-20T04:35:05Z

With this PR, on 1.11 (because SnoopCompile is broken on julia master)

julia> using SnoopCompileCore

julia> invs = @snoop_invalidations using HTTP;

julia> using SnoopCompile, AbstractTrees

julia> trees = invalidation_trees(invs)
8-element Vector{SnoopCompile.MethodInvalidations}:
 inserting joinpath(uri::URIs.URI, parts::String...) @ URIs ~/.julia/packages/URIs/7Tg5r/src/URIs.jl:584 invalidated:
   mt_backedges: 1: signature Tuple{typeof(joinpath), Any, String} triggered MethodInstance for Artifacts.jointail(::Any, ::String) (0 children)

 inserting write(io::IO, s::Union{Base.AnnotatedString, SubString{<:Base.AnnotatedString}}) @ StyledStrings ~/.julia/juliaup/julia-1.11.1+0.aarch64.apple.darwin14/share/julia/stdlib/v1.11/StyledStrings/src/io.jl:255 invalidated:
   backedges: 1: superseding write(io::IO, s::AbstractString) @ Base strings/io.jl:196 with MethodInstance for write(::HTTP.Streams.Stream{HTTP.Messages.Response, HTTP.Connections.Connection{Sockets.TCPSocket}}, ::Base.AnnotatedString{String}) (1 children)

 inserting write(io::IO, c::Base.AnnotatedChar) @ StyledStrings ~/.julia/juliaup/julia-1.11.1+0.aarch64.apple.darwin14/share/julia/stdlib/v1.11/StyledStrings/src/io.jl:270 invalidated:
   backedges: 1: superseding write(io::IO, x) @ Base io.jl:792 with MethodInstance for write(::HTTP.Streams.Stream{HTTP.Messages.Response, HTTP.Connections.Connection{Sockets.TCPSocket}}, ::Base.AnnotatedChar{Char}) (1 children)

 inserting print(io::IO, c::Base.AnnotatedChar) @ StyledStrings ~/.julia/juliaup/julia-1.11.1+0.aarch64.apple.darwin14/share/julia/stdlib/v1.11/StyledStrings/src/io.jl:281 invalidated:
   backedges: 1: superseding print(io::IO, c::AbstractChar) @ Base char.jl:253 with MethodInstance for print(::IOContext{IOBuffer}, ::Base.AnnotatedChar{Char}) (1 children)
              2: superseding print(io::IO, c::AbstractChar) @ Base char.jl:253 with MethodInstance for print(::IOBuffer, ::AbstractChar) (1 children)
              3: superseding print(io::IO, c::AbstractChar) @ Base char.jl:253 with MethodInstance for print(::Base.AnnotatedIOBuffer, ::Base.AnnotatedChar{Char}) (1 children)

 inserting show(io::IO, c::Base.AnnotatedChar) @ StyledStrings ~/.julia/juliaup/julia-1.11.1+0.aarch64.apple.darwin14/share/julia/stdlib/v1.11/StyledStrings/src/io.jl:283 invalidated:
   backedges: 1: superseding show(io::IO, c::AbstractChar) @ Base char.jl:283 with MethodInstance for show(::IOBuffer, ::AbstractChar) (12 children)

 inserting typeinfo_implicit(::Type{Dates.Second}) @ Dates ~/.julia/juliaup/julia-1.11.1+0.aarch64.apple.darwin14/share/julia/stdlib/v1.11/Dates/src/periods.jl:23 invalidated:
   backedges: 1: superseding typeinfo_implicit(T) @ Base arrayshow.jl:550 with MethodInstance for Base.typeinfo_implicit(::Any) (22 children)
   1 mt_cache

 inserting kwcall(::NamedTuple, ::typeof(printstyled), io::Base.AnnotatedIOBuffer, msg...) @ StyledStrings.Legacy ~/.julia/juliaup/julia-1.11.1+0.aarch64.apple.darwin14/share/julia/stdlib/v1.11/StyledStrings/src/legacy.jl:126 invalidated:
   backedges: 1: superseding kwcall(::NamedTuple, ::typeof(printstyled), io::IO, msg...) @ Base util.jl:141 with MethodInstance for Core.kwcall(::@NamedTuple{bold::Bool, italic::Bool, underline::Bool, blink::Bool, reverse::Bool, hidden::Bool, color::Symbol}, ::typeof(printstyled), ::IO, ::String, ::String, ::String) (1 children)
              2: superseding kwcall(::NamedTuple, ::typeof(printstyled), io::IO, msg...) @ Base util.jl:141 with MethodInstance for Core.kwcall(::@NamedTuple{bold::Bool, italic::Bool, underline::Bool, blink::Bool, reverse::Bool, hidden::Bool, color::Symbol}, ::typeof(printstyled), ::IO, ::String, ::String) (1 children)
              3: superseding kwcall(::NamedTuple, ::typeof(printstyled), io::IO, msg...) @ Base util.jl:141 with MethodInstance for Core.kwcall(::@NamedTuple{bold::Bool, italic::Bool, underline::Bool, blink::Bool, reverse::Bool, hidden::Bool, color::Symbol}, ::typeof(printstyled), ::IO, ::String, ::Vararg{String}) (15 children)
              4: superseding kwcall(::NamedTuple, ::typeof(printstyled), io::IO, msg...) @ Base util.jl:141 with MethodInstance for Core.kwcall(::@NamedTuple{bold::Bool, italic::Bool, underline::Bool, blink::Bool, reverse::Bool, hidden::Bool, color::Int64}, ::typeof(printstyled), ::IO, ::String, ::String) (1 children)
              5: superseding kwcall(::NamedTuple, ::typeof(printstyled), io::IO, msg...) @ Base util.jl:141 with MethodInstance for Core.kwcall(::@NamedTuple{bold::Bool, italic::Bool, underline::Bool, blink::Bool, reverse::Bool, hidden::Bool, color::Symbol}, ::typeof(printstyled), ::IO, ::String) (6 children)
              6: superseding kwcall(::NamedTuple, ::typeof(printstyled), io::IO, msg...) @ Base util.jl:141 with MethodInstance for Core.kwcall(::@NamedTuple{bold::Bool, italic::Bool, underline::Bool, blink::Bool, reverse::Bool, hidden::Bool, color::Int64}, ::typeof(printstyled), ::IO, ::String) (1 children)

 inserting print(io::IO, s::Union{Base.AnnotatedString, SubString{<:Base.AnnotatedString}}) @ StyledStrings ~/.julia/juliaup/julia-1.11.1+0.aarch64.apple.darwin14/share/julia/stdlib/v1.11/StyledStrings/src/io.jl:258 invalidated:
   backedges: 1: superseding print(io::IO, s::AbstractString) @ Base strings/io.jl:195 with MethodInstance for print(::IOContext{IOBuffer}, ::Base.AnnotatedString{String}) (56 children)

Sixzero · 2024-11-20T10:56:16Z

Really nice, how fast this got fixed!

Removing TTFX for every julia session for simple .get request is really a valuable thing in my opinion.

Going to test the solution now!

Sixzero · 2024-11-20T11:15:50Z

Okay, we are actually faster than CurlHTTP.jl! Also it fixed post request times!

This is insane! I don't know if precompilation of all julia packages took longer, but this is lifechanging, finally get/post requests are fast in julia!

IanButterworth · 2024-11-20T15:03:37Z

@quinnj perhaps you're best positioned to review this?

src/precompile.jl

quinnj

Yeah, I think this looks pretty great.

IanButterworth · 2024-11-20T17:33:04Z

Great. What about the __init__'s?

I think they're robust to being called twice in a row.. i.e. push!(empty!
Except for ConnectionRequest.__init__() which just pushes? Perhaps that one's not needed here?
https://github.com/IanButterworth/HTTP.jl/blob/4510bd2a0cb818abe71d75fb71d2b57c01b0dfa4/src/clientlayers/ConnectionRequest.jl#L23-L31

Sixzero · 2024-11-21T13:29:38Z

Things work well after 1 day of usage. I guess compat is also good.

IanButterworth · 2024-11-21T13:37:44Z

@quinnj I don't have merge rights here

timholy · 2024-11-22T09:13:54Z

Calling __init__ from a precompile workload should be fine. The TTL increase is a little sad. (We really should start caching the results of invalidation checks.) Is there any part of this workload that you don't need precompiled and could be moved to a @setup block?

src/precompile.jl

KristofferC · 2024-11-22T10:40:27Z

The time to load increase makes the total latency bigger than it was before, or?

Regarding inits, those should already run, no?

IanButterworth · 2024-11-22T11:21:31Z

Time to load is the same. I showed time to precompile.

KristofferC · 2024-11-22T11:23:18Z

Aha, the using made me think it was just loading. All good then 👍. (I see @timholy made the same mistake heh).

fredrikekre · 2024-11-22T13:11:02Z

It is interesting that this works so well given that the test case takes the https path while precompilation is for http.

fredrikekre · 2024-11-22T13:12:22Z

julia> using HTTP

julia> @time HTTP.get("http://google.com");
  1.416276 seconds (1.73 M allocations: 83.683 MiB, 2.23% gc time, 66.31% compilation time)

julia> using HTTP

julia> @time HTTP.get("https://google.com");
  2.386149 seconds (3.09 M allocations: 153.946 MiB, 2.69% gc time, 95.02% compilation time)

fredrikekre · 2024-11-22T13:24:15Z

Note also that HTTP compiles quite a lot depending on the input arguments, but this PR is still very benefitial for these cases too:

julia> using HTTP

julia> @time HTTP.get("https://google.com");
  2.792014 seconds (2.88 M allocations: 152.468 MiB, 1.08% gc time, 93.55% compilation time: 35% of which was recompilation)

julia> @time HTTP.get("https://google.com", ["User-Agent" => "julia"]);
  0.722978 seconds (34.91 k allocations: 1.862 MiB, 3 lock conflicts, 11.12% compilation time)

julia> @time HTTP.get("https://google.com"; retries = 5);
  0.440146 seconds (266.69 k allocations: 13.969 MiB, 3 lock conflicts, 78.77% compilation time)

julia> @time HTTP.get("https://google.com"; retry = false);
  0.447537 seconds (263.32 k allocations: 13.759 MiB, 3 lock conflicts, 77.86% compilation time)

julia> @time HTTP.get("https://google.com"; redirect_limit = 5);
  0.170655 seconds (43.77 k allocations: 2.449 MiB, 3 lock conflicts, 39.95% compilation time)

IanButterworth · 2024-11-22T14:07:57Z

I know you're saying that its probably not necessary, but if I bake those in

julia> using HTTP
Precompiling HTTP...
Info Given HTTP was explicitly requested, output will be shown live
  2.267022 seconds (15.86 M allocations: 800.601 MiB, 4.71% gc time, 99.81% compilation time)
  0.020154 seconds (125.76 k allocations: 6.660 MiB, 96.36% compilation time)
  0.075205 seconds (334.63 k allocations: 17.019 MiB, 99.14% compilation time)
  0.075720 seconds (331.12 k allocations: 16.825 MiB, 99.05% compilation time)
  0.023594 seconds (108.50 k allocations: 5.781 MiB, 97.43% compilation time)
  1 dependency successfully precompiled in 7 seconds. 29 already precompiled.
  1 dependency had output during precompilation:
┌ HTTP
│  [Output was shown above]
└

julia> begin
           url = "https://google.com";
           @time HTTP.get(url);
           @time HTTP.get(url, ["User-Agent" => "julia"]);
           @time HTTP.get(url; retries = 5);
           @time HTTP.get(url; retry = false);
           @time HTTP.get(url; redirect_limit = 5);
       end;
  1.208905 seconds (3.10 M allocations: 154.791 MiB, 3.29% gc time, 84.32% compilation time)
  0.092161 seconds (13.77 k allocations: 853.656 KiB, 2 lock conflicts, 6.37% compilation time)
  0.094209 seconds (673 allocations: 185.328 KiB, 2 lock conflicts)
  0.097040 seconds (653 allocations: 255.398 KiB, 2 lock conflicts)
  0.095959 seconds (629 allocations: 183.906 KiB, 2 lock conflicts)

Without baking the kwarg ones in

  1.187503 seconds (3.10 M allocations: 154.793 MiB, 3.23% gc time, 85.06% compilation time)
  0.124479 seconds (134.01 k allocations: 6.998 MiB, 2 lock conflicts, 31.16% compilation time)
  0.346849 seconds (338.27 k allocations: 17.200 MiB, 0.98% gc time, 2 lock conflicts, 66.06% compilation time)
  0.313539 seconds (328.23 k allocations: 16.698 MiB, 1.03% gc time, 2 lock conflicts, 67.35% compilation time)
  0.148446 seconds (106.91 k allocations: 5.685 MiB, 2 lock conflicts, 29.42% compilation time)

Worth it?

Also, I assume it's not trivial to cover https in this offline workload?

fredrikekre · 2024-11-22T14:24:49Z

I don't think it is worth it, I just picked some keywords at random, and there are lots of them.

Also, I assume it's not trivial to cover https in this offline workload?

You can probably do something like

HTTP.jl/test/client.jl

Line 511 in b7f5aa8

# Set up server with self-signed cert

. If that is succesful it is probably not worth precompiling the http path.

IanButterworth · 2024-11-22T14:45:03Z

Great, that works.

Now

julia> begin
               url = "https://google.com"
               @time HTTP.get(url);
               @time HTTP.get(url, ["User-Agent" => "julia"]);
               @time HTTP.get(url; retries = 5);
               @time HTTP.get(url; retry = false);
               @time HTTP.get(url; redirect_limit = 5);
       end;
  0.725483 seconds (1.74 M allocations: 83.781 MiB, 8.31% gc time, 70.41% compilation time)
  0.126748 seconds (134.04 k allocations: 6.999 MiB, 1.92% gc time, 2 lock conflicts, 34.05% compilation time)
  0.351043 seconds (338.45 k allocations: 17.206 MiB, 2 lock conflicts, 70.20% compilation time)
  0.322628 seconds (328.27 k allocations: 16.698 MiB, 1.44% gc time, 2 lock conflicts, 68.70% compilation time)
  0.151559 seconds (106.95 k allocations: 5.688 MiB, 2 lock conflicts, 34.52% compilation time)

IanButterworth · 2024-11-22T15:17:30Z

New nightly failures are some unexpected allocations

HTTPVersion: Test Failed at /home/runner/work/HTTP.jl/HTTP.jl/test/httpversion.jl:19
  Expression: #= /home/runner/work/HTTP.jl/HTTP.jl/test/httpversion.jl:19 =# @allocated(HTTPVersion("1.1")) == 0
   Evaluated: 16 == 0

  Expression: #= /home/runner/work/HTTP.jl/HTTP.jl/test/httpversion.jl:20 =# @allocated(VersionNumber(HTTPVersion("1.1"))) == 0
   Evaluated: 64 == 0

IanButterworth · 2024-11-23T03:52:34Z

Merge?

quinnj · 2024-11-23T04:16:00Z

src/HTTP.jl

@@ -631,4 +631,9 @@ function Base.parse(::Type{T}, str::AbstractString)::T where T <: Message
    return m
 end

+# only run if precompiling pkgimages
+if VERSION >= v"1.9.0-0" && Base.JLOptions().use_pkgimages != 0 && ccall(:jl_generating_output, Cint, ()) == 1


Yeesh; it'd be nice if PrecompileTools provided an API for all these conditions (isprecompiling?)

quinnj · 2024-11-23T04:16:58Z

src/precompile.jl

+    # random port in the dynamic/private range (49152–65535) which are are
+    # least likely to be used by well-known services
+    _port = 57813
+


Suggested change

# random port in the dynamic/private range (49152–65535) which are are

# least likely to be used by well-known services

_port = 57813

This broke things because _port is an arg given to serve!, but surprisingly not 1.10, which I don't understand...

btw, I had tried do listenany without a port arg and got a method error.

I've reverted the change. And I think the reason 1.10 didn't fail is that I think on 1.10 if coverage is requested pkgimages aren't used.

So I just made the precompile workload run whether or not pkgimages are requested, so that coverage isn't weird in tests.

quinnj

Thanks for doing this @IanButterworth!

This reverts commit cb9b1f2.

timholy · 2024-11-23T13:07:05Z

Late comment:

I don't think it is worth it, I just picked some keywords at random, and there are lots of them.

Typically, the answer to this is to add Base.@constprop :none in front of the function. Unless there's really some reason we think it's important to compile for each separate combination of boolean or symbol keyword arguments.

IanButterworth · 2024-11-23T16:10:49Z

Thanks. I took a look into adding that but it's not clear where to, and my attempts were in vain. There's a lot of functions returning functions, so it's not particularly clear where to put the annotation.

@quinnj says he is close to the 2.0 rewrite being ready, so it might not be worth it, unless it jumps out to someone.

timholy · 2024-11-24T12:04:54Z

Gotcha. I assumed it was compiling many variants of get, but perhaps @fredrikekre's examples are instead meant to exercise additional code paths that are hidden behind runtime dispatch.

The use-case for @constprop :none is if you end up doing multiple expensive compiles for specific arg/kwarg combinations. Functions that take Int, Bool, and Symbol args are particularly at-risk for "dumb" recompiles. You can use MethodAnalysis.jl to count the number of MethodInstances for a specific function. This is also a use-case for pgds which lets you check a whole workload.

This should probably be used more heavily in packages than it is.

DilumAluthge · 2024-11-25T01:08:56Z

I suspect that this PR broke precompilation of HTTP on Julia 1.9. See for example: https://github.com/JuliaRegistries/General/actions/runs/12001290587/job/33451587371

Can we skip the precompile workload on Julia 1.9?

IanButterworth · 2024-11-25T20:46:06Z

Just a trailing note.. I was underestimating the improvement here because it turns out google.com is quite slow to resolve for me...

julia> using HTTP

julia> @time HTTP.get("https://julialang.org");
  0.052138 seconds (2.42 k allocations: 368.070 KiB, 49.11% compilation time)

julia> @time HTTP.get("https://julialang.org");
  0.009134 seconds (292 allocations: 212.633 KiB, 1 lock conflict)

whereas before this PR

julia> using HTTP

julia> @time HTTP.get("https://julialang.org");
  4.739799 seconds (14.82 M allocations: 740.448 MiB, 2.34% gc time, 99.51% compilation time)

julia> @time HTTP.get("https://julialang.org");
  0.009363 seconds (288 allocations: 212.633 KiB, 1 lock conflict)

Sixzero · 2024-11-27T21:46:29Z

Love to see! Insane improvements guys!

krynju · 2024-11-28T16:27:21Z

didn't see anyone showing first server response time changes

#### 1.10.10
server = HTTP.serve!("127.0.0.1", 8088; ) do req
    HTTP.Response(200)
end


julia> using HTTP

julia> server = HTTP.serve!("127.0.0.1", 8088; ) do req
           HTTP.Response(200)
       end
[ Info: Listening on: 127.0.0.1:8088, thread id: 1

shell> time curl localhost:8088
curl 'localhost:8088'  0.00s user 0.01s system 0% cpu 1.325 total

shell> time curl localhost:8088
curl 'localhost:8088'  0.01s user 0.00s system 91% cpu 0.006 total


#### 1.10.12

julia> using HTTP

julia> server = HTTP.serve!("127.0.0.1", 8088; ) do req
           HTTP.Response(200)
       end
[ Info: Listening on: 127.0.0.1:8088, thread id: 1

shell> time curl localhost:8088
curl 'localhost:8088'  0.00s user 0.00s system 0% cpu 0.479 total

shell> time curl localhost:8088
curl 'localhost:8088'  0.00s user 0.01s system 84% cpu 0.007 total

IanButterworth mentioned this pull request Nov 20, 2024

TTFX time of request is high. MVP #1194

Closed

add an offline precompile workload

86fb137

IanButterworth force-pushed the ib/precompile branch from dba39cc to 86fb137 Compare November 20, 2024 03:49

quinnj reviewed Nov 20, 2024

View reviewed changes

src/precompile.jl Outdated Show resolved Hide resolved

quinnj approved these changes Nov 20, 2024

View reviewed changes

suggestions

4510bd2

IanButterworth added 2 commits November 20, 2024 12:36

fix

4f3ecfc

tweak inits

72075d1

KristofferC reviewed Nov 22, 2024

View reviewed changes

src/precompile.jl Show resolved Hide resolved

precompile https

9f82f5b

only run workload if precompiling pkgimages

b3e27a7

IanButterworth force-pushed the ib/precompile branch from d03303d to b3e27a7 Compare November 22, 2024 14:57

Merge branch 'master' into ib/precompile

62d8390

quinnj reviewed Nov 23, 2024

View reviewed changes

Update src/precompile.jl

cb9b1f2

quinnj approved these changes Nov 23, 2024

View reviewed changes

IanButterworth added 2 commits November 22, 2024 23:37

Revert "Update src/precompile.jl"

b4358cd

This reverts commit cb9b1f2.

precompile wheter or not pkgimages are used

7ae48cd

quinnj merged commit 6aa6981 into JuliaWeb:master Nov 23, 2024
10 of 13 checks passed

IanButterworth deleted the ib/precompile branch November 23, 2024 05:01

DilumAluthge mentioned this pull request Nov 25, 2024

Don't run the precompile workload on Julia 1.9 #1199

Closed

This was referenced Nov 25, 2024

Fix trailing task precompilation issue on 1.9. Run CI on 1.9 #1201

Merged

Precompilation spits out Task object, followed by segfaults during runtime, on 1.9 #1202

Closed

IanButterworth mentioned this pull request Dec 12, 2024

Reduce the time-to-first-request by adding pregenerated precompile statements #805

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

add an offline precompile workload #1195

add an offline precompile workload #1195

IanButterworth commented Nov 20, 2024 •

edited

Loading

IanButterworth commented Nov 20, 2024 •

edited

Loading

Sixzero commented Nov 20, 2024

Sixzero commented Nov 20, 2024

IanButterworth commented Nov 20, 2024

quinnj left a comment

IanButterworth commented Nov 20, 2024

Sixzero commented Nov 21, 2024 •

edited

Loading

IanButterworth commented Nov 21, 2024

timholy commented Nov 22, 2024

KristofferC commented Nov 22, 2024

IanButterworth commented Nov 22, 2024

KristofferC commented Nov 22, 2024

fredrikekre commented Nov 22, 2024

fredrikekre commented Nov 22, 2024

fredrikekre commented Nov 22, 2024

IanButterworth commented Nov 22, 2024 •

edited

Loading

fredrikekre commented Nov 22, 2024

IanButterworth commented Nov 22, 2024

IanButterworth commented Nov 22, 2024

IanButterworth commented Nov 23, 2024

quinnj Nov 23, 2024

quinnj Nov 23, 2024

IanButterworth Nov 23, 2024

IanButterworth Nov 23, 2024

quinnj left a comment

timholy commented Nov 23, 2024

IanButterworth commented Nov 23, 2024

timholy commented Nov 24, 2024

DilumAluthge commented Nov 25, 2024

IanButterworth commented Nov 25, 2024

Sixzero commented Nov 27, 2024

krynju commented Nov 28, 2024

	# random port in the dynamic/private range (49152–65535) which are are
	# least likely to be used by well-known services
	_port = 57813

add an offline precompile workload #1195

add an offline precompile workload #1195

Conversation

IanButterworth commented Nov 20, 2024 • edited Loading

IanButterworth commented Nov 20, 2024 • edited Loading

Sixzero commented Nov 20, 2024

Sixzero commented Nov 20, 2024

IanButterworth commented Nov 20, 2024

quinnj left a comment

Choose a reason for hiding this comment

IanButterworth commented Nov 20, 2024

Sixzero commented Nov 21, 2024 • edited Loading

IanButterworth commented Nov 21, 2024

timholy commented Nov 22, 2024

KristofferC commented Nov 22, 2024

IanButterworth commented Nov 22, 2024

KristofferC commented Nov 22, 2024

fredrikekre commented Nov 22, 2024

fredrikekre commented Nov 22, 2024

fredrikekre commented Nov 22, 2024

IanButterworth commented Nov 22, 2024 • edited Loading

fredrikekre commented Nov 22, 2024

IanButterworth commented Nov 22, 2024

IanButterworth commented Nov 22, 2024

IanButterworth commented Nov 23, 2024

quinnj Nov 23, 2024

Choose a reason for hiding this comment

quinnj Nov 23, 2024

Choose a reason for hiding this comment

IanButterworth Nov 23, 2024

Choose a reason for hiding this comment

IanButterworth Nov 23, 2024

Choose a reason for hiding this comment

quinnj left a comment

Choose a reason for hiding this comment

timholy commented Nov 23, 2024

IanButterworth commented Nov 23, 2024

timholy commented Nov 24, 2024

DilumAluthge commented Nov 25, 2024

IanButterworth commented Nov 25, 2024

Sixzero commented Nov 27, 2024

krynju commented Nov 28, 2024

IanButterworth commented Nov 20, 2024 •

edited

Loading

IanButterworth commented Nov 20, 2024 •

edited

Loading

Sixzero commented Nov 21, 2024 •

edited

Loading

IanButterworth commented Nov 22, 2024 •

edited

Loading