Since its inception rustbuild has always worked in three stages: one for
libstd, one for libtest, and one for rustc. These three stages were
architected around crates.io dependencies, where rustc wants to depend
on crates.io crates but said crates don't explicitly depend on libstd,
requiring a sysroot assembly step in the middle. This same logic was
applied for libtest where libtest wants to depend on crates.io crates
(`getopts`) but `getopts` didn't say that it depended on std, so it
needed `std` built ahead of time.
Lots of time has passed since the inception of rustbuild, however,
and we've since gotten to the point where even `std` itself is depending
on crates.io crates (albeit with some wonky configuration). This
commit applies the same logic to the two dependencies that the `test`
crate pulls in from crates.io, `getopts` and `unicode-width`. Over the
many years since rustbuild's inception `unicode-width` was the only
dependency picked up by the `test` crate, so the extra configuration
necessary to get crates building in this crate graph is unlikely to be
too much of a burden on developers.
After this patch it means that there are now only two build phasese of
rustbuild, one for libstd and one for rustc. The libtest/libproc_macro
build phase is all lumped into one now with `std`.
This was originally motivated by rust-lang/cargo#7216 where Cargo was
having to deal with synthesizing dependency edges but this commit makes
them explicit in this repository.
Enable thumbv7a-pc-windows-msvc target build end to end in rust/master
With this PR, plus another commit https://github.com/rust-lang-nursery/compiler-builtins/commit/cf98161da7ed5217b6031796f0f60b4dd07148a4, I'm able to build the target thumbv7a-pc-windows-msvc successfully, and I'm able to use the stage2 artifacts to build arm32 projects. The commit in compiler_builtins is in release 0.1.14, the current cargo.lock in rust master still uses 0.1.12, so I bumped the compiler_builtins version in cargo.lock to 0.1.15
The command I used to build rust
```
c:\python27\python.exe x.py build --host x86_64-pc-windows-msvc --build x86_64-pc-windows-msvc --target thumbv7a-pc-windows-msvc --verbose
```
**Changes**
1. update cargolock to use compiler_builtins 0.1.15
2. handle libunwind in libtest for thumv7a the same as what we have for aarch64
3. in llvm codegen add a field in CodegenContext to carry the arch, so later in create_msvc_imps function, the arch can be used to check against "x86", instead of 32 pointer width. Apparently Thumv7a is handled differently than x86.
**Background**
I'm from Microsoft working on enabling Azure IoTEdge on ARM32 Windows IoTCore, Azure IoTEdge has a component called IoTEdged written in rust as a NT service running on Windows, so we need to enable rust on thumbv7a in order to have full IoTEdge. My colleague had made some heavy lifting and we've been using our private toolchain to build IoTEdged in our devops pipeline, because at that time we cannot build thumbv7a target end to end successfully. This change is a followup to enable the end to end build for thumbv7a-pc-windows-msvc target.
**Next step**
I'll submit more changes to have this target built nightly in rust/master, to achieve the same availability for aarch64-pc-windows-msvc, indexed here https://rust-lang.github.io/rustup-components-history/aarch64-pc-windows-msvc.html and can be manually installed. **Please do share what takes to make this happen, is there a formal process I need to follow\?**
This changes removes a cyclic dependency between the "test" and "libtest"
crates, where "libtest" depends on "test" for "black_box", but "test" depends on
"libtest" for everything else.
I've chosen the "hint" module because there seems to be enough consensus in the
discussion of RFC2360 that this module is where such an intrinsic would belong,
but this PR does not implement that RFC! (note: if that RFC ever gets merged,
the API, docs, etc. of this API will need to change).
For backwards compatibility reasons I've chosen to also keep the "test" feature
gate for these instead of adding a new feature gate. If we change the feature
gate, we'll potentially all benchmarks, and while that's something that we could
do, it seems unnecessary to do that now - if RFC2360 gets merged, we'll need to
do that anyways.
libtest: Use deterministic HashMap, avoid spawning thread if there is no concurrency
It seems desirable to make a test and bench runner deterministic, which this achieves by using a deterministic hasher. Also, we we only have 1 thread, we don't bother spawning one and just use the main thread.
The motivation for this is to be able to run the test harness in miri, where we can neither access the OS RNG, nor spawn threads.
Alex Crichton