Currently our slowest test suite on android, run-pass, takes over 5 times longer
than the x86_64 component (~400 -> ~2200s). Typically QEMU emulation does indeed
add overhead, but not 5x for this kind of workload. One of the slowest parts of
the Android process is that *compilation* happens serially. Tests themselves
need to run single-threaded on the emulator (due to how the test harness works)
and this forces the compiles themselves to be single threaded.
Now Travis gives us more than one core per machine, so it'd be much better if we
could take advantage of them! The emulator itself is still fundamentally
single-threaded, but we should see a nice speedup by sending binaries for it to
run much more quickly.
It turns out that we've already got all the tools to do this in-tree. The
qemu-test-{server,client} that are in use for the ARM Linux testing are a
perfect match for the Android emulator. This commit migrates the custom adb
management code in compiletest/rustbuild to the same qemu-test-{server,client}
implementation that ARM Linux uses.
This allows us to lift the parallelism restriction on the compiletest test
suites, namely run-pass. Consequently although we'll still basically run the
tests themselves in single threaded mode we'll be able to compile all of them in
parallel, keeping the pipeline much more full and using more cores for the work
at hand. Additionally the architecture here should be a bit speedier as it
should have less overhead than adb which is a whole new process on both the host
and the emulator!
Locally on an 8 core machine I've seen the run-pass test suite speed up from
taking nearly an hour to only taking 6 minutes. I don't think we'll see quite a
drastic speedup on Travis but I'm hoping this change can place the Android tests
well below 2 hours instead of just above 2 hours.
Because the client/server here are now repurposed for more than just QEMU,
they've been renamed to `remote-test-{server,client}`.
Note that this PR does not currently modify how debuginfo tests are executed on
Android. While parallelizable it wouldn't be quite as easy, so that's left to
another day. Thankfully that test suite is much smaller than the run-pass test
suite.
As a final fix I discovered that the ARM and Android test suites were actually
running all library unit tests (e.g. stdtest, coretest, etc) twice. I've
corrected that to only run tests once which should also give a nice boost in
overall cycle time here.
Run tests for the cargo submodule in tree
Previously the `cargotest` suite would run some arbitrary revision of Cargo's
test suite, but now that we're bundling it in tree we should be running the
Cargo submodule's test suite instead.
LLVM 4.0 Upgrade
Since nobody has done this yet, I decided to get things started:
**Todo:**
* [x] push the relevant commits to `rust-lang/llvm` and `rust-lang/compiler-rt`
* [x] cleanup `.gitmodules`
* [x] Verify if there are any other commits from `rust-lang/llvm` which need backporting
* [x] Investigate / fix debuginfo ("`<optimized out>`") failures
* [x] Use correct emscripten version in docker image
---
Closes#37609.
---
**Test results:**
Everything is green 🎉
Previously the `cargotest` suite would run some arbitrary revision of Cargo's
test suite, but now that we're bundling it in tree we should be running the
Cargo submodule's test suite instead.
Currently the Cargo binary has jumped from 14M to 34M on the beta channel, which
appears to be due to the fact that we're compiling tools with debug information
inside them. This additionally means that the `rls` binary is 62M right now!
This wasn't an intentional change, so be sure to disable debuginfo when
compiling tools as it's just intended for the standard library and compile for
now.
* Use the right version when building combined installer
* Update dependencies of rls as it depends on rustc and plugins
* Fix build-manifest and the versions it uses for the rls
Use rls-data crate
This basically pulls out a bunch of data structures used by save-analysis for serialization into an external crate, and pulls that crate in using Rustbuild. The RLS can then share these data structures with the compiler which in some cases will allow more efficient communication between the compiler and the RLS (i.e., without serialisation).
Along the way, I have to pull in rls-span, which is the RLS's way of defining spans (more type-safe than the compiler's built-in way). This is basically just to convert from compiler spans to RLS spans.
I also pull in the crates.io version of rustc-serialize, which is a bit annoying, but seems to be the only way to have serialisable data in an external crate. To make this work, all of the save-analysis crate has to use this version too (cc #40527).
Finally I pull in a line from #40347 to make the unstable crate checking stuff working.
There are a lot of changes to save-analysis but they are all mechanical and trivial - changing from using `From` to `Into` (because of orphan rules) being the main thing.
r? @alexcrichton
rustbuild: Make save-analysis an option
This makes save-analysis an option independent from the release channel.
The CI build scripts have been modified to enable the flag.
*Merge with caution.* I haven't tested this, and this can cause nightly breakage.
This commit updates Cargo with rust-lang/cargo#3820 which includes a fix for
rust-lang/cargo#3819. At the same time this also slightly tweaks how rustbuild
builds cargo to ensure that all the build information (including git info and
such) makes its way into the binary.
Closesrust-lang/cargo#3820
rustbuild: Build documentation for `proc_macro`
This commit fixes#38749 by building documentation for the `proc_macro` crate by
default for configured hosts. Unfortunately did not turn out to be a trivial
fix. Currently rustbuild generates documentation into multiple locations: one
for std, one for test, and one for rustc. The initial fix for this issue simply
actually executed `cargo doc -p proc_macro` which was otherwise completely
elided before.
Unfortunately rustbuild was the left to merge two documentation trees together.
One for the standard library and one for the rustc tree (which only had docs for
the `proc_macro` crate). Rustdoc itself knows how to merge documentation files
(specifically around search indexes, etc) but rustbuild was unaware of this, so
an initial fix ended up destroying the sidebar and the search bar from the
libstd docs.
To solve this issue the method of documentation has been tweaked slightly in
rustbuild. The build system will not use symlinks (or directory junctions on
Windows) to generate all documentation into the same location initially. This'll
rely on rustdoc's logic to weave together all the output and ensure that it ends
up all consistent.
Closes#38749
This causes problems when first cloning and bootstrapping the repository
unfortunately, so let's ensure that Cargo sticks around in its own workspace.
Because Cargo is a submodule it's not available by default on the inital clone
of the rust-lang/rust repository. Normally it's the responsibility of the
rustbuild to take care of this, but unfortunately to build rustbuild itself we
need to resolve the workspace conflicts.
To deal with this we'll just have to ensure that all submodules are in their own
workspace, which sort of makes sense anyway as updates to dependencies as
bugfixes to Cargo should go to rust-lang/cargo instead of rust-lang/rust. In any
case this commit removes Cargo from the global workspace which should resolve
the issues that we've been seeing.
To actually perform this the `cargo` submodule has been moved to the top
directory to ensure it's outside the scope of `src/Cargo.toml` as a workspace.
This commit fixes#38749 by building documentation for the `proc_macro` crate by
default for configured hosts. Unfortunately did not turn out to be a trivial
fix. Currently rustbuild generates documentation into multiple locations: one
for std, one for test, and one for rustc. The initial fix for this issue simply
actually executed `cargo doc -p proc_macro` which was otherwise completely
elided before.
Unfortunately rustbuild was the left to merge two documentation trees together.
One for the standard library and one for the rustc tree (which only had docs for
the `proc_macro` crate). Rustdoc itself knows how to merge documentation files
(specifically around search indexes, etc) but rustbuild was unaware of this, so
an initial fix ended up destroying the sidebar and the search bar from the
libstd docs.
To solve this issue the method of documentation has been tweaked slightly in
rustbuild. The build system will not use symlinks (or directory junctions on
Windows) to generate all documentation into the same location initially. This'll
rely on rustdoc's logic to weave together all the output and ensure that it ends
up all consistent.
Closes#38749
This commit adds support to rustbuild for compiling Cargo as part of the release
process. Previously rustbuild would simply download a Cargo snapshot and
repackage it. With this change we should be able to turn off artifacts from the
rust-lang/cargo repository and purely rely on the artifacts Cargo produces here.
The infrastructure added here is intended to be extensible to other components,
such as the RLS. It won't exactly be a one-line addition, but the addition of
Cargo didn't require too much hooplah anyway.
The process for release Cargo will now look like:
* The rust-lang/rust repository has a Cargo submodule which is used to build a
Cargo to pair with the rust-lang/rust release
* Periodically we'll update the cargo submodule as necessary on rust-lang/rust's
master branch
* When branching beta we'll create a new branch of Cargo (as we do today), and
the first commit to the beta branch will be to update the Cargo submodule to
this exact revision.
* When branching stable, we'll ensure that the Cargo submodule is updated and
then make a stable release.
Backports to Cargo will look like:
* Send a PR to cargo's master branch
* Send a PR to cargo's release branch (e.g. rust-1.16.0)
* Send a PR to rust-lang/rust's beta branch updating the submodule
* Eventually send a PR to rust-lang/rust's master branch updating the submodule
For reference, the process to add a new component to the rust-lang/rust release
would look like:
* Add `$foo` as a submodule in `src/tools`
* Add a `tool-$foo` step which compiles `$foo` with the specified compiler,
likely mirroring what Cargo does.
* Add a `dist-$foo` step which uses `src/tools/$foo` and the `tool-$foo` output
to create a rust-installer package for `$foo` likely mirroring what Cargo
does.
* Update the `dist-extended` step with a new dependency on `dist-$foo`
* Update `src/tools/build-manifest` for the new component.
travis: Compile a more compatible libc.a for musl
The mitigations for #34978 involve passing `-Wa,-mrelax-relocations=no` to all C
code we compile, and we just forgot to pass it when compiling musl itself.
Closes#39979
This is a simple way to workaround the debugging issues caused by the rustc
wrapper used in the bootstrap process. Namely, it uses some obscure environment
variables and you can’t just copy the failed command and run it in the shell or
debugger to examine the failure more closely.
With `--on-fail` its possible to run an arbitrary command within exactly the
same environment under which rustc failed. Theres’s multiple ways to use this
new flag:
$ python x.py build --stage=1 --on-fail=env
would print a list of environment variables and the failed command, so a
few copy-pastes and you now can run the same rust in your shell outside the
bootstrap system.
$ python x.py build --stage=1 --on-fail=bash
Is a more useful variation of the command above in that it launches a whole
shell with environment already in place! All that’s left to do is copy-paste
the command just above the shell prompt!
Fixes#38686Fixes#38221
Add support for test suites emulated in QEMU
This commit adds support to the build system to execute test suites that cannot
run natively but can instead run inside of a QEMU emulator. A proof-of-concept
builder was added for the `arm-unknown-linux-gnueabihf` target to show off how
this might work.
In general the architecture is to have a server running inside of the emulator
which a local client connects to. The protocol between the server/client
supports compiling tests on the host and running them on the target inside the
emulator.
Closes#33114
This commit adds support to the build system to execute test suites that cannot
run natively but can instead run inside of a QEMU emulator. A proof-of-concept
builder was added for the `arm-unknown-linux-gnueabihf` target to show off how
this might work.
In general the architecture is to have a server running inside of the emulator
which a local client connects to. The protocol between the server/client
supports compiling tests on the host and running them on the target inside the
emulator.
Closes#33114
This commit adds a new flag to the configure script,
`--enable-extended`, which is intended for specifying a desire to
compile the full suite of Rust tools such as Cargo, the RLS, etc. This
is also an indication that the build system should create combined
installers such as the pkg/exe/msi artifacts.
Currently the `--enable-extended` flag just indicates that combined
installers should be built, and Cargo is itself not compiled just yet
but rather only downloaded from its location. The intention here is to
quickly get to feature parity with the current release process and then
we can start improving it afterwards.
All new files in this PR inside `src/etc/installer` are copied from the
rust-packaging repository.
This commit changes all tools and such to get compiled in stage0, not in
later stages. The purpose of this commit is to cut down dependencies on later
stages for future modifications to the build system. Notably we're going to be
adding builders that produce a full suite of cross-compiled artifacts for a
particular host, and that shouldn't compile the `x86_64-unknown-linux-gnu`
compiler more than once. Currently dependencies on, for example, the error index
end up compiling the `x86_64-unknown-linux-gnu` compiler more than necessary.
As a result here we move many dependencies on these tools to being produced by a
stage0 compiler, not a stage1+ compiler. None of these tools actually need to be
staged at all, so they'll exhibit consistent behavior across the stages.
This commit switches the rustbuild build system to compiling the
compiler twice for a normal bootstrap rather than the historical three
times.
Rust is a bootstrapped language which means that a previous version of
the compiler is used to build the next version of the compiler. Over
time, however, we change many parts of compiler artifacts such as the
metadata format, symbol names, etc. These changes make artifacts from
one compiler incompatible from another compiler. Consequently if a
compiler wants to be able to use some artifacts then it itself must have
compiled the artifacts.
Historically the rustc build system has achieved this by compiling the
compiler three times:
* An older compiler (stage0) is downloaded to kick off the chain.
* This compiler now compiles a new compiler (stage1)
* The stage1 compiler then compiles another compiler (stage2)
* Finally, the stage2 compiler needs libraries to link against, so it
compiles all the libraries again.
This entire process amounts in compiling the compiler three times.
Additionally, this process always guarantees that the Rust source tree
can compile itself because the stage2 compiler (created by a freshly
created compiler) would successfully compile itself again. This
property, ensuring Rust can compile itself, is quite important!
In general, though, this third compilation is not required for general
purpose development on the compiler. The third compiler (stage2) can
reuse the libraries that were created during the second compile. In
other words, the second compilation can produce both a compiler and the
libraries that compiler will use. These artifacts *must* be compatible
due to the way plugins work today anyway, and they were created by the
same source code so they *should* be compatible as well.
So given all that, this commit switches the default build process to
only compile the compiler three times, avoiding this third compilation
by copying artifacts from the previous one. Along the way a new entry in
the Travis matrix was also added to ensure that our full bootstrap can
succeed. This entry does not run tests, though, as it should not be
necessary.
To restore the old behavior of a full bootstrap (three compiles) you can
either pass:
./configure --enable-full-bootstrap
or if you're using config.toml:
[build]
full-bootstrap = true
Overall this will hopefully be an easy 33% win in build times of the
compiler. If we do 33% less work we should be 33% faster! This in turn
should affect cycle times and such on Travis and AppVeyor positively as
well as making it easier to work on the compiler itself.
This commit switches the default build system for Rust from the makefiles to
rustbuild. The rustbuild build system has been in development for almost a year
now and has become quite mature over time. This commit is an implementation of
the proposal on [internals] which slates deletion of the makefiles on
2016-01-02.
[internals]: https://internals.rust-lang.org/t/proposal-for-promoting-rustbuild-to-official-status/4368
This commit also updates various documentation in `README.md`,
`CONTRIBUTING.md`, `src/bootstrap/README.md`, and throughout the source code of
rustbuild itself.
Closes#37858
