Openembedded Core Discussions
 help / color / mirror / Atom feed
From: "S. Lockwood-Childs" <sjl@vctlabs.com>
To: openembedded-core@lists.openembedded.org
Subject: Re: [PATCH] rust: fix arm64 link failures when building rust apps
Date: Thu, 16 Dec 2021 15:04:30 -0800	[thread overview]
Message-ID: <20211216230430.GF5920@codepurple> (raw)
In-Reply-To: <20211216091940.GE5920@codepurple>

Patch withdrawn: turns out meta-tegra was pulling in an old toolchain
for CUDA reasons, and that just got fixed. Probably there isn't another
aarch64 platform still using a non-fixed toolchain, and if there happens
to be one maybe they will run across this patch to carry it in their own
layer.

Note: apologies for not CC-ing Alexander Kanavin, but I couldn't figure
out how to directly reply to his question from this list's web UI & I don't
have a copy of his email in my inbox.
https://lists.openembedded.org/g/openembedded-core/message/159775

On Thu, Dec 16, 2021 at 01:19:45AM -0800, S. Lockwood-Childs wrote:
> Typical error when trying to build a rust app (for example, librsvg)
> for aarch64 targets looks like:
>   undefined references to `__aarch64_ldadd8_rel'
> 
> The upstream rust commit
> "add target feature outline-atomics" 0f9f241aac21bc77fb9e757da18207abefdc841d
> has caused a number of such link failure regressions on platforms with aarch64
> toolchains that are not able to cope with -moutline-atomics flag. This includes
> musl toolchains [1] but some glibc toolchains also are not able to handle this
> flag [2,3]
> 
> OE gcc-cross-aarch64 is currently one of the latter, but when it *is* able to
> handle outline-atomics this patch should go away, to take advantage of the
> reported performance benefit.
> 
> [1] https://github.com/rust-lang/git2-rs/issues/706
> [2] https://www.mail-archive.com/gcc-bugs@gcc.gnu.org/msg661738.html
> [3] https://bugzilla.redhat.com/show_bug.cgi?id=1830472
> 
> Signed-off-by: S. Lockwood-Childs <sjl@vctlabs.com>
> ---
>  meta/recipes-devtools/rust/rust-target.inc    |  2 +
>  .../rust/rust-disable-outline-atomics.patch   | 50 +++++++++++++++++++
>  2 files changed, 52 insertions(+)
>  create mode 100644 meta/recipes-devtools/rust/rust/rust-disable-outline-atomics.patch
> 
> diff --git a/meta/recipes-devtools/rust/rust-target.inc b/meta/recipes-devtools/rust/rust-target.inc
> index 3f637b3ba5..bac743f66d 100644
> --- a/meta/recipes-devtools/rust/rust-target.inc
> +++ b/meta/recipes-devtools/rust/rust-target.inc
> @@ -2,6 +2,8 @@ require rust.inc
>  
>  DEPENDS += "rust-llvm (=${PV})"
>  
> +SRC_URI += "file://rust-disable-outline-atomics.patch"
> +
>  # Otherwise we'll depend on what we provide
>  INHIBIT_DEFAULT_RUST_DEPS:class-native = "1"
>  # We don't need to depend on gcc-native because yocto assumes it exists
> diff --git a/meta/recipes-devtools/rust/rust/rust-disable-outline-atomics.patch b/meta/recipes-devtools/rust/rust/rust-disable-outline-atomics.patch
> new file mode 100644
> index 0000000000..fc99a06a6a
> --- /dev/null
> +++ b/meta/recipes-devtools/rust/rust/rust-disable-outline-atomics.patch
> @@ -0,0 +1,50 @@
> +rust: fix arm64 link failures when building rust apps
> +
> +Typical error when trying to build a rust app looks like: 
> +  undefined references to `__aarch64_ldadd8_rel'
> +
> +The upstream rust commit 
> +"add target feature outline-atomics" 0f9f241aac21bc77fb9e757da18207abefdc841d
> +has caused a number of such link failure regressions on platforms with aarch64
> +toolchains that are not able to cope with -moutline-atomics flag. This includes
> +musl toolchains [1] but some glibc toolchains also are not able to handle this
> +flag [2,3]
> +
> +OE gcc-cross-aarch64 is currently one of the latter, but when it *is* able to
> +handle outline-atomics this patch should go away, to take advantage of the
> +reported performance benefit.
> +
> +[1] https://github.com/rust-lang/git2-rs/issues/706
> +[2] https://www.mail-archive.com/gcc-bugs@gcc.gnu.org/msg661738.html
> +[3] https://bugzilla.redhat.com/show_bug.cgi?id=1830472
> +
> +Upstream-Status: Inappropriate [other]
> +
> +Signed-off-by: S. Lockwood-Childs <sjl@vctlabs.com>
> +
> +diff --git a/compiler/rustc_codegen_llvm/src/llvm_util.rs b/compiler/rustc_codegen_llvm/src/llvm_util.rs
> +index c2136f161..ca221eb4a 100644
> +--- a/compiler/rustc_codegen_llvm/src/llvm_util.rs
> ++++ b/compiler/rustc_codegen_llvm/src/llvm_util.rs
> +@@ -416,13 +416,14 @@ pub fn llvm_global_features(sess: &Session) -> Vec<String> {
> +     // -Ctarget-features
> +     features.extend(sess.opts.cg.target_feature.split(',').flat_map(&filter));
> + 
> +-    // FIXME: Move outline-atomics to target definition when earliest supported LLVM is 12.
> +-    if get_version() >= (12, 0, 0)
> +-        && sess.target.llvm_target.contains("aarch64-unknown-linux")
> +-        && sess.target.llvm_target != "aarch64-unknown-linux-musl"
> +-    {
> +-        features.push("+outline-atomics".to_string());
> +-    }
> ++//    // FIXME: Move outline-atomics to target definition when earliest supported LLVM is 12.
> ++//    if get_version() >= (12, 0, 0)
> ++//        && sess.target.llvm_target.contains("aarch64-unknown-linux")
> ++//        && sess.target.llvm_target != "aarch64-unknown-linux-musl"
> ++//    {
> ++//        features.push("+outline-atomics".to_string());
> ++//    }
> ++
> + 
> +     features
> + }
> -- 
> 2.20.1
> 


  parent reply	other threads:[~2021-12-16 23:04 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-12-16  9:19 [PATCH] rust: fix arm64 link failures when building rust apps S. Lockwood-Childs
2021-12-16 10:06 ` [OE-core] " Alexander Kanavin
2021-12-16 23:04 ` S. Lockwood-Childs [this message]
2021-12-17 13:11   ` Alex Kiernan

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20211216230430.GF5920@codepurple \
    --to=sjl@vctlabs.com \
    --cc=openembedded-core@lists.openembedded.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox