We'll use this as a temporary measure to get an LLVM update landed, but
we'll have to go through and update images later to make sure they've
got the right toolchains.
ci: fix docker cache hash collision
#58416 uncovered a bug in our caching for docker images: if the image `foo` pulls files from the image `bar` and a file in `bar` changed, the hash of `foo` will be the same even though it should be different. In that PR's case, `dist-i686-linux` pulls scripts from `dist-x86_64-linux`, and the PR only changed those scripts, causing an hash collision for `dist-i686-linux`.
We have to fix this, since the image will be rebuilt every time bors switches from testing master to testing beta/stable (and when it switches back), making CI way more painful than it currently is.
The approach used by this PR is to just include all the files in `src/ci/docker` in the hash. It's a bit heavy-handed and it will cause a rebuild of all the images every time a single image changes, but it's the best I can think of.
r? @Mark-Simulacrum
cc @alexcrichton @kennytm
Before this commit the hash used to cache docker images was calculated
from the image's files and the files in the scripts/ directory. This
worked fine when all the files used by an image were in those
directories, but some images pull files from other images, causing hash
collisions in some cases.
This commit changes the hash to include the files of all the docker
images, causing a rebuild of all the images when a single one changes.
That's a bit heavy-handed, but we have no way to track which files an
image pulls in and hash collisions are really painful to deal with.
[CI] Update GCC in the dist-linux builders
Upstream LLVM is planning to raise their minimum toolchain requirements, so they may start using C++14 features. This new policy has already landed in the form of a "soft" error.
http://lists.llvm.org/pipermail/llvm-dev/2019-February/130117.html
For GCC, they will require at least version 5.1. This PR moves our crosstool-ng builders to their max GCC 5.2, with a few small patches to fix compatibility. The dist-x86_64-linux builder is updated to GCC 5.5 and LLVM/Clang 8.0.0-rc2, which also affects dist-i686-linux sharing the same scripts.
r? @alexcrichton
The Ubuntu keyserver is more reliable than the MIT PGP server, which is
prone to going down. This commit also explicitly uses port 80 on the
keyserver for reasons outlined in #57844.
NVPTX target specification
This change adds a built-in `nvptx64-nvidia-cuda` GPGPU no-std target specification and a basic PTX assembly smoke tests.
The approach is taken here and the target spec is based on `ptx-linker`, a project started about 1.5 years ago. Key feature: bitcode object files being linked with LTO into the final module on the linker's side.
Prior to this change, the linker used a `ld` linker-flavor, but I think, having the special CLI convention is a more reliable way.
Questions about further progress on reliable CUDA workflow with Rust:
1. Is it possible to create a test suite `codegen-asm` to verify end-to-end integration with LLVM backend?
1. How would it be better to organise no-std `compile-fail` tests: add `#![no_std]` where possible and mark others as `ignore-nvptx` directive, or alternatively, introduce `compile-fail-no-std` test suite?
1. Can we have the `ptx-linker` eventually be integrated as `rls` or `clippy`? Hopefully, this should allow to statically link against LLVM used in Rust and get rid of the [current hacky solution](https://github.com/denzp/rustc-llvm-proxy).
1. Am I missing some methods from `rustc_codegen_ssa:🔙:linker::Linker` that can be useful for bitcode-only linking?
Currently, there are no major public CUDA projects written in Rust I'm aware of, but I'm expecting to have a built-in target will create a solid foundation for further experiments and awesome crates.
Related to #38789Fixes#38787Fixes#38786
compiletest: Support opt-in Clang-based run-make tests and use them for testing xLTO.
Some cross-language run-make tests need a Clang compiler that matches the LLVM version of `rustc`. Since such a compiler usually isn't available these tests (marked with the `needs-matching-clang`
directive) are ignored by default.
For some CI jobs we do need these tests to run unconditionally though. In order to support this a `--force-clang-based-tests` flag is added to compiletest. If this flag is specified, `compiletest` will fail if it can't detect an appropriate version of Clang.
@rust-lang/infra The PR doesn't yet enable the tests yet. Do you have any recommendation for which jobs to enable them?
cc #57438
r? @alexcrichton
Attempt to debug 259 exit code on AppVeyor
Let's try to dig in a bit more and see where this is coming from, it
looks like AppVeyor is also unsure where this is coming from!
Build the standard library for thumbv7neon-unknown-linux-gnueabihf in CI
Using the `dist-armv7-linux` image instead of `dist-various-1` in order to use the ARMv7 toolchain available in `dist-armv7-linux`.
Closes#57030.
The new git submodule src/llvm-project is a monorepo replacing src/llvm
and src/tools/{clang,lld,lldb}. This also serves as a rebase for these
projects to the new 8.x branch from trunk.
The src/llvm-emscripten fork is unchanged for now.
Upgrade x86_64-fortanix-unknown-sgx platform support to tier 2
## Overview
1. This PR upgrades x86_64-fortanix-unknown-sgx platform support to tier 2 (std only) by setting up build automation for this target.
1. For supporting unwinding, this target needs to link to a port of LLVM's libunwind (more details could be found in #56979), which will be distributed along with the Rust binaries (similar to the extra musl objects)
### Building and copying libunwind:
We have added a new build script (`build-x86_64-fortanix-unknown-sgx-toolchain.sh`) that will run while the container is built. This will build `libunwind.a` from git source.
While the container is built, the persistent volumes where obj/ gets created aren't yet mapped. As a workaround, we copy the built `libunwind.a` to `obj/build/x86_64-unknown-linux-gnu/stage2/lib/rustlib/x86_64-fortanix-unknown-sgx/lib/` after x.py runs.
If any reviewer knows of a better solution, please do tell.
r? @Mark-Simulacrum
The version included in Ubuntu 16.04 repositories in the dist-various-1
docker, Arm GCC version 4.9, does not support the new Armv8-M
architecture.
This commit adds the team-gcc-arm-embedded PPA to get through APT a
newer version of Arm GCC.
