Test cross compilation with proc-macro crates
In particular, test #37899.
Can you actually somehow distinguish between target and host on proc-macro crates? What if you want to implement a feature for one given target architecture, independent from whether it gets created by cross compilation or normal compilation? How do build.rs scripts handle this?
est31 at 2016-11-25 05:35:39
@est31 I think we shouldn't support cross compiling
proc-macrocrates -- that is, we always compileproc-macrocrates for the host architecture. This issue is calling for testing of cross compiled crates that depend onproc-macrocrates (which are compiled for the host) as well as other dependencies (which are compiled for the target).Jeffrey Seyfried at 2016-11-25 05:47:35
I've posted a stripped-down example program which triggers this error.
Eric Kidd at 2016-11-26 13:38:25
This seems like a E-needstest, probably sufficient to just have a run-make test I guess, which will then be tested on the cross test platforms. @jseyfried Does that sound correct? Please untag as needstest if that's wrong.
Mark Rousskov at 2017-05-16 15:45:27
@Mark-Simulacrum Yeah, that sounds correct -- a run-make test is probably best here.
Jeffrey Seyfried at 2017-05-16 20:20:37
Triage: not aware of anything here changing, but I'm also not an expert.
Steve Klabnik at 2020-07-11 20:57:58