Failure to build Soss/MonteCarloMeasurements in Quickstart

[sethaxen]

Currently, the docs take nearly 2 hours to build for some reason, nearly all of that time spent on the Quickstart. The Quickstart currently errors while compiling Soss, causing everything afterward to fail. I suspect that that failure is related to the massive slowdown. The error however seems to come from precompiling MonteCarloMeasurements. After calling using Soss, I get the following stacktrace (See for example this build):

ERROR: LoadError: LoadError: UndefVarError: PT not defined
Stacktrace:
 [1] top-level scope at /home/vsts/.julia/packages/MonteCarloMeasurements/bf9PV/src/particles.jl:122
 [2] include at ./boot.jl:328 [inlined]
 [3] include_relative(::Module, ::String) at ./loading.jl:1105
 [4] include at ./Base.jl:31 [inlined]
 [5] include(::String) at /home/vsts/.julia/packages/MonteCarloMeasurements/bf9PV/src/MonteCarloMeasurements.jl:1
 [6] top-level scope at /home/vsts/.julia/packages/MonteCarloMeasurements/bf9PV/src/MonteCarloMeasurements.jl:79
 [7] include at ./boot.jl:328 [inlined]
 [8] include_relative(::Module, ::String) at ./loading.jl:1105
 [9] include(::Module, ::String) at ./Base.jl:31
 [10] top-level scope at none:2
 [11] eval at ./boot.jl:330 [inlined]
 [12] eval(::Expr) at ./client.jl:425
 [13] top-level scope at ./none:3
in expression starting at /home/vsts/.julia/packages/MonteCarloMeasurements/bf9PV/src/particles.jl:122
in expression starting at /home/vsts/.julia/packages/MonteCarloMeasurements/bf9PV/src/MonteCarloMeasurements.jl:79
ERROR: LoadError: Failed to precompile MonteCarloMeasurements [0987c9cc-fe09-11e8-30f0-b96dd679fdca] to /home/vsts/.julia/compiled/v1.3/MonteCarloMeasurements/tHYBD_7eNvK.ji.
Stacktrace:
 [1] error(::String) at ./error.jl:33
 [2] compilecache(::Base.PkgId, ::String) at ./loading.jl:1283
 [3] _require(::Base.PkgId) at ./loading.jl:1024
 [4] require(::Base.PkgId) at ./loading.jl:922
 [5] require(::Module, ::Symbol) at ./loading.jl:917
 [6] include at ./boot.jl:328 [inlined]
 [7] include_relative(::Module, ::String) at ./loading.jl:1105
 [8] include(::Module, ::String) at ./Base.jl:31
 [9] top-level scope at none:2
 [10] eval at ./boot.jl:330 [inlined]
 [11] eval(::Expr) at ./client.jl:425
 [12] top-level scope at ./none:3
in expression starting at /home/vsts/.julia/packages/Soss/2R1Bp/src/Soss.jl:12
┌ Warning: failed to run `@setup` block in src/quickstart.md

@cscherrer and @baggepinnen, have you seen errors like this before, or do you have any idea what could be causing them?

This particular build used Soss v0.10.0 and MonteCarloMeasurements v0.6.4.

[baggepinnen]

Hmm, I haven't seen any massive compilation time like that, full tests for MCM usually complete in about 14 minutes on Travis and has done for the last couple of months. PT is the variable used in most code generation in MCM, so unless Soss does its own code generation for particles it does sound like the problem is coming from MCM. How does it look if you run using a later version of MCM?

[sethaxen]

Hmm, I haven't seen any massive compilation time like that, full tests for MCM usually complete in about 14 minutes on Travis and has done for the last couple of months. PT is the variable used in most code generation in MCM, so unless Soss does its own code generation for particles it does sound like the problem is coming from MCM. How does it look if you run using a later version of MCM?

So I did some troubleshooting, and I think these are two separate issues.

The PT error logged above is eliminated with newer versions of MonteCarloMeasurements, which we can use if we pin Soss#master; the next Soss release will bump the compat.

The slowdown can be replicated by installing MCMCChains and Soss in a clean environment and executing

import MCMCChains # this is fast
using Soss # this takes a very long time

Reversing the order removes the slowdown. So this is either a Soss or MCMCChains issue, and it happens even on Soss#master. I'll close this and open a new Soss issue.

Thanks for taking a look at it!

[baggepinnen]

MCMC chains is known to cause this issue with turing as well, I think there is a similar issue on the turing repo, or possibly on turing slack.

[sethaxen]

Yes, there's this issue, but I'm not positive it's related: TuringLang/MCMCChains.jl#171. It will be easier to tell once MCMCChains has had its refactor.