disable autoprecompile by default

[IanButterworth]

With native code caching incoming, doing (auto)precompilation at this stage of CI is suboptimal because the native caches will be generated for a julia setup that doesn't match the test julia setup i.e. bounds checking. which would cause precompilation in this step and in a following julia-runtest step.

This disables auto-precompilation here, but exposes it as an option to re-enable.

My hope here is that we can release this as a non-breaking change so we can avoid a load more unnecessary precompilation once native caching lands.

[codecov-commenter]

Codecov Report

Base: 100.00% // Head: 100.00% // No change to project coverage 👍

Coverage data is based on head (64b7e04) compared to base (f995fa4).
Patch has no changes to coverable lines.

Additional details and impacted files

@@            Coverage Diff            @@
##            master       #26   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            1         1           
  Lines            2         2           
=========================================
  Hits             2         2           

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[IanButterworth]

@SaschaMann if this requires your sign off, the thread here has more context JuliaLang/Pkg.jl#3281

[SaschaMann]

My hope here is that we can release this as a non-breaking change so we can avoid a load more unnecessary precompilation once native caching lands.

Do you happen to know a scenario where this could be breaking?

I'm fine with merging it as it is though.

[IanButterworth]

I don't think so. This step promises to build the package, the fact it autoprecompiled wasn't really within the spec I think

[kleinschmidt]

FWIW this did actually cause breakage in some internal code for us but in a VERY specific way: during CI, a different project is instantiated (with test dependencies), and that gets precompiled during Pkg.test(). But we have some code that uses Distributed.LocalManager to add additional workers with addprocs(; exeflags="--project") which in production is The Right Thing to do but during tests results in a precompilation race condition which causes mysterious segfaults/compilecache misses in dependencies; before, the "extra" precompilation that julia-buildpkg was doing meant that the project was appropriately precompiled and there was no race condition. The fix was to just activate the actual project that is active on teh manager when addproces is called, instead of relying on exeflags="--project". Again, I don't think this is really the FAULT of this change, but wanted to raise it just in case it's worth knowing for anyone else :)

[kleinschmidt]

All of which is to say https://xkcd.com/1172/