The target specs that rustc will use won't be relevant to the build system. Adding things to platform.mk is how the build system is extended, right now, and I don't foresee that changing.
On Fri, Oct 17, 2014 at 5:19 PM, Cody P Schafer <d...@codyps.com> wrote: > On Fri, Oct 17, 2014 at 4:41 PM, Ben Harris <m...@bharr.is> wrote: >> I don't believe it implies that. Only rustc would be interpreting the json >> file (first the snapshot rustc, then the stage0 rustc). >> >> Are any variables in the JSON relevant to configure or make? I thought only >> the target would be needed, and that is passed as a variable. >> > > My impression was that the linker (cc) was relevent to the initial > build of the included libraries. If all of the calls to target build > tools are driven by rustc (and never driven directly by > make/configure), then it will be fine. > > If make or configure ever call the linker (or any other target build > tools), it will need to parse the target JSON. > > Looking at the build process, at least jemalloc (which I've run into > bugs cross building) needs the target build tools (not just the > linker). > > As a result, we'd need to parse json before building jemalloc (pretty > sure rustc isn't calling that build) > > Also, I'd have to check to be sure, but I think the build of jemalloc > will need more than a "linker". Even assuming the linker is also the C > compiler (as is the current case) might be insufficient. This means > the linker json currently proposed will need a few more fields to be > useful for the initial build/configure. > _______________________________________________ > Rust-dev mailing list > Rust-dev@mozilla.org > https://mail.mozilla.org/listinfo/rust-dev -- http://octayn.net/ _______________________________________________ Rust-dev mailing list Rust-dev@mozilla.org https://mail.mozilla.org/listinfo/rust-dev
