Closed warricksothr closed 8 years ago
This is a recent development, so I think it might have to do with using the older compiler. The same issue does not happen on the more modern ARMv7 container.
So I'm going to compile and new cmake and new clang in the ARMv6 container and try to use those to build the Rust binaries.
Compiling a new cmake and clang didn't resolve the issue. Looks like it might be a far more simple case of a llvm recipe that shouldn't exist anymore. Attempting a build after a proper cleaning of the required directory.
I built Clang 3.8 so I think I'll keep that around for the builds as long as it produces usable binaries with the correct glibc.
1.5.0 Nightly just built on Raspbian with Clang 3.8. Need to test this on a Raspberry Pi to confirm that it is behaving as expected.
Success!!! Switched to Clang 3.7, crosscompiled LLVM and Clang and we're now up and running again. New Raspbian builds will be built on Clang 3.7 as it seems GCC 4.8 will no longer build Rust reliably.
pi@raspberrypi ~/dev/hello_world $ rustc -vV
rustc 1.5.0-dev (e50298f7b 2015-10-07)
binary: rustc
commit-hash: e50298f7b53d4f51c936c0f61e282d6cb7ff2a19
commit-date: 2015-10-07
host: arm-unknown-linux-gnueabihf
release: 1.5.0-dev
pi@raspberrypi ~/dev/hello_world $ cargo -V
cargo 0.6.0-nightly (1e5fa84 2015-10-08)
pi@raspberrypi ~/dev/hello_world $ cargo clean
pi@raspberrypi ~/dev/hello_world $ cargo build
Compiling hello_world v0.1.0 (file:///home/pi/dev/hello_world)
pi@raspberrypi ~/dev/hello_world $ cargo run
Running `target/debug/hello_world`
Hello, world!
pi@raspberrypi ~/dev/hello_world $
Going to let the autobuild process take over this again.
Latest stack trace from the Raspbian container.