A basic LLVM 12 build provides clang-12 but not clang++-12, but
it does provide both clang and clang++, so we shouldn't hard fail
when first checking for clang-12 and clang++-12.
Feature no-patches is ever only useful when setting other env variables
BORING_BSSL{,_FIPS}{,_SOURCE}_PATH, and it has no impact on the APIs
provided by any of the boring crates, so we may as well make it an env
variable itself so downstream users have less features to propagate
across their own crate graph.
Builds using feature fips or fips-link-precompiled now
read variables prefixed by BORING_BSSL_FIPS_ instead of
BORING_BSSL_. This helps complex builds where build dependencies
also use boring, where we may not want to use fips there.
Without those separate variables, the boring build for the
build dependencies end up relying on e.g. BORING_BSSL_PATH,
causing errors if this path is a boring checkout intended for
fips builds, while the fips feature isn't enabled for
the build dependency.
This means BORING_SSL_PRECOMPILED_BCM_O is now
BORING_BSSL_PRECOMPILED_BCM_O.
Prefix BORING_BSSL_ has been chosen because that's the
one that is used the most among all the variables
the build script uses.
Using a struct improves navigation of the build script,
as we can rely on rust-analyzer to help us check how
a feature flag or an environment variable is used,
as opposed to grepping for multiple env::var calls
or #[cfg] attributes.
This commit also removes some obsolete blocks of code
related to the now defunct ndk-old-gcc and fuzzing features.
Overwrite boringSSL's default key exchange preferences with safe
defaults using feature flags:
* "kx-pq-supported" enables support for PQ key exchange algorithms.
Classical key exchange is still preferred, but will be upgraded to PQ
if requested.
* "kx-pq-preferred" enables preference for PQ key exchange,
with fallback to classical key exchange if requested.
* "kx-nist-required" disables non-NIST key exchange.
Each feature implies "kx-safe-default". When this feature is enabled,
don't compile bindings for `SSL_CTX_set1_curves()` and `SslCurve`. This
is to prevent the feature flags from silently overriding curve
preferences chosen by the user.
Ideally we'd allow both: that is, use "kx-*" to set defaults, but still
allow the user to manually override them. However, this doesn't work
because by the time the `SSL_CTX` is constructed, we don't yet know
whether we're the client or server. (The "kx-*" features set different
preferences for each.) If "kx-sfe-default" is set, then the curve
preferences are set just before initiating a TLS handshake
(`SslStreamBuilder::connect()`) or waiting for a TLS handshake
(`SslStreamBuilder::accept()`).
This commit modifies the Cargo `include` field for `boring-sys` to
include all the files necessary to actually build the FIPS-certified
revision of BoringSSL. Currently, some of these files are missing (see
#157 for details on this).
This branch improves on my previous approach in PR #158, which switched
from using a Cargo `include` to a Cargo `exclude`. Using `exclude`
rather than `include` resulted in a much larger crates.io package, but
at the time, I thought this was less likely to result in breakage in the
future, because I was concerned about the inability to verify that the
set of excludes/includes can build a new pinned `boringssl` git revision
without having to actually publish a crates.io release.
However, as @nox pointed out in [this comment][1], `cargo package` can
be used to verify a build with the `exclude`s/`include`s applied. This
branch therefore adds `cargo package` steps to CI that check that the
package can actually be built. This way, we are able to make a much
smaller change to the included files, resulting in a smaller package
published to crates.io.
On this branch, the package is 6.7MiB compressed, which is not much
larger than it was previously:
```
Finished dev [unoptimized + debuginfo] target(s) in 55.65s
Packaged 1851 files, 33.7MiB (6.7MiB compressed)
```
Fixes#157Closes#158
[1]: https://github.com/cloudflare/boring/pull/158#issuecomment-1693067112,
This was originally going to be fixed by #101, however that PR was closed and superseded by #117, which was missing this fix.
The original problem was caused by #97, which updated boringssl to a version that included [a change that removed hmac.h from ssl.h](05b360d797).
This PR adds an include for hmac.h, so it is again available through boring-sys.
Previously we were building from the deps directory with submodules. For publishing we were copying files in sumbodules into the package. With this we were making the package directory dirty with build artifacts and applied patches.
This commit change the build script's behaviour: sources are now copied to the output directory and then boringssl is built from there.
In addition, this commit adds files that were missing from the package for building with patches.
This longer path (inside the prebuilt toolchain included in the NDK)
has been the preferred sysroot since NDK r19. Newer NDKs no longer
have a top-level "sysroot" directory at all.
Cross-compiling to AArch64 Linux can be done with a CMake toolchain
file, along with setting the correct compiler and include paths in the
environment.
Cross-compiling from X64 Windows to ARM64 Windows doesn't look at the
toolchain at all, because CMake + Visual Studio can already
cross-compile. Unfortunately, the Visual Studio CMake generator
doesn't set CMAKE_SYSTEM_PROCESSOR, which is what the BoringSSL
CMakeLists.txt is looking at to choose the architecture. For now,
disable the use of assembly when cross-compiling on Windows (assuming
that the Visual Studio generator will be used there).
As pointed out in the comment, bindgen generates tests that cause
compiler warnings about misaligned references. bindgen people are
aware of the issue, but we have to deal with our warnings that are
treated as errors. For the time being, suppress alignment tests
on platforms that are known to be triggering UB.
I suspect that other non-x86 platforms are affected as well, but I can't
get the tests to compile for those tests at the moment, so I'm not sure.
Dealing with the issues one platform at a time.
cfg!() is evaluated for the host OS executing build.rs script.
What we need here is to look whether we are building *for* macOS.
Otherwise, for example, builds for iOS on macOS will try to add this
flag, causing warnings since rustc does not build cdylibs on iOS.
When bindgen generates bindings for iOS, it must be told to use iOS
sysroot with all the standard C headers. Otherwise it tries using
the host macOS headers and fails miserably.
Anthony Ramine