Regression in rust nightly 2017-03-04 #40493
Assignees
Labels
P-high
High priority
regression-from-stable-to-beta
Performance or correctness regression from stable to beta.
T-compiler
Relevant to the compiler team, which will review and decide on the PR/issue.
Comments
arielb1
added
regression-from-stable-to-nightly
|
1.17 is now beta |
arielb1
added
the
T-compiler
|
Probably a duplicate of #40573 -- stack usage jumped up for some reason |
|
triage: P-high |
arielb1
added a commit
to arielb1/rust
that referenced
this issue
Apr 2, 2017
LLVM has a bug - PR32488 - where it fails to deduplicate allocas in some circumstances. The function `start_new_block` has allocas totalling 1216 bytes, and when LLVM inlines several copies of that function into the recursive function `expr::into`, that function's stack space usage goes into tens of kiBs, causing stack overflows. Mark `start_new_block` as inline(never) to keep it from being inlined, getting stack usage under control. Fixes rust-lang#40493. Fixes rust-lang#40573.
arielb1
added a commit
to arielb1/rust
that referenced
this issue
Apr 2, 2017
LLVM has a bug - PR32488 - where it fails to deduplicate allocas in some circumstances. The function `start_new_block` has allocas totalling 1216 bytes, and when LLVM inlines several copies of that function into the recursive function `expr::into`, that function's stack space usage goes into tens of kiBs, causing stack overflows. Mark `start_new_block` as inline(never) to keep it from being inlined, getting stack usage under control. Fixes rust-lang#40493. Fixes rust-lang#40573.
|
Fixes are enqueued. Need confirmation from op yet. |
|
@brson Trying to get this patched branch build and tested right now. I will report back asap. |
|
|
|
Excellent. |
arielb1
pushed a commit
to arielb1/rust
that referenced
this issue
Apr 5, 2017
…hton mark build::cfg::start_new_block as inline(never) LLVM has a bug - [PR32488](https://bugs.llvm.org//show_bug.cgi?id=32488) - where it fails to deduplicate allocas in some circumstances. The function `start_new_block` has allocas totalling 1216 bytes, and when LLVM inlines several copies of that function into the recursive function `expr::into`, that function's stack space usage goes into tens of kiBs, causing stack overflows. Mark `start_new_block` as inline(never) to keep it from being inlined, getting stack usage under control. Fixes rust-lang#40493. Fixes rust-lang#40573. r? @eddyb
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I cannot compile my library
wlc(on crates.io) on nightlies later then 2017-03-03 anymore.The code is not using any unstable features, if compiled with the feature "unsafe-stable" and results do not differ between "default" features (which uses
optin_builtin_traits) and the "unsafe-stable" feature (which does disableoptin_builtin_traits).So code, that still works on stable and beta, does not compile anymore without any helpful error messages (see below).
Overview:
Working:
rustup run stable cargo build --features "unsafe-stable"rustup run beta cargo build --features "unsafe-stable"rustup run nightly-2017-03-03 cargo build --features "unsafe-stable"Broken:
rustup run nightly-2017-03-04 cargo build --features "unsafe-stable"rustup run nightly-2017-03-04 cargo buildBoth abort with a stack overflow:
rustup run nightly-2017-03-12 cargo build --features "unsafe-stable"rustup run nightly-2017-03-12 cargo buildBoth fail without any specific error:
I have no idea, where I should start to debug this. Reproducing this error is also quite annoying to setup, as the library depends on a bunch of native libraries not available in most repositories. So I am thankful for any hints about how I could provide more information.
All tests where done after running
cargo clean.Meta
rustc --version --verbose:The text was updated successfully, but these errors were encountered: