Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

DWARF info for static vars in lib crates has stopped being produced reliably in LTO builds #90357

Closed
Tracked by #129080
cbiffle opened this issue Oct 28, 2021 · 7 comments · Fixed by #95685
Closed
Tracked by #129080
Assignees
Labels
A-debuginfo Area: Debugging information in compiled programs (DWARF, PDB, etc.) A-LLVM Area: Code generation parts specific to LLVM. Both correctness bugs and optimization-related issues. A-lto Area: Link Time Optimization A-reproducibility Area: Reproducible / Deterministic builds C-bug Category: This is a bug. P-high High priority regression-from-stable-to-stable Performance or correctness regression from one stable version to another. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue.
Milestone

Comments

@cbiffle
Copy link
Contributor

cbiffle commented Oct 28, 2021

As of the 2021-09-23 nightly, rustc has stopped producing debug output for variables in a specific combination of circumstances:

  • The variable is static
  • It is defined in a lib crate.
  • The lib crate is included in a bin crate.
  • The bin crate is built with lto = true.

We noticed this after bumping past 09-23 to the 2021-10-27 nightly, when our debugger broke. The static is still being included in the ELF output (both allocated space, and represented correctly in the symtab), but is missing from DWARF. Bisecting indicates 2021-09-23 as the culprit (09-22 works).

Here is a repro case that I've reduced that works on linux x86-64, at least, and probably other DWARF platforms.

https://github.com/cbiffle/rustc-dwarf-regression-repro

@rustbot modify labels: +regression-from-stable-to-nightly -regression-untriaged

@cbiffle cbiffle added C-bug Category: This is a bug. regression-untriaged Untriaged performance or correctness regression. labels Oct 28, 2021
@rustbot rustbot added I-prioritize Issue: Indicates that prioritization has been requested for this issue. regression-from-stable-to-nightly Performance or correctness regression from stable to nightly. and removed regression-untriaged Untriaged performance or correctness regression. labels Oct 28, 2021
@cbiffle cbiffle changed the title DWARF info for static vars in lib crates has stopped being produced reliably DWARF info for static vars in lib crates has stopped being produced reliably in LTO builds Oct 28, 2021
@camelid camelid added A-debuginfo Area: Debugging information in compiled programs (DWARF, PDB, etc.) E-needs-bisection Call for participation: This issue needs bisection: https://github.com/rust-lang/cargo-bisect-rustc labels Oct 28, 2021
cbiffle added a commit to oxidecomputer/hubris that referenced this issue Oct 28, 2021
This is the last toolchain that does not suffer from this bug:

rust-lang/rust#90357
cbiffle added a commit to oxidecomputer/hubris that referenced this issue Oct 28, 2021
This is the last toolchain that does not suffer from this bug:

rust-lang/rust#90357
@Mark-Simulacrum
Copy link
Member

Bumping to regression-stable-to-beta, because 2021-09-23 is a 1.57 nightly. But would be good to confirm this reproduces on beta.

@Mark-Simulacrum Mark-Simulacrum added this to the 1.57.0 milestone Oct 28, 2021
@Mark-Simulacrum Mark-Simulacrum added regression-from-stable-to-beta Performance or correctness regression from stable to beta. and removed regression-from-stable-to-nightly Performance or correctness regression from stable to nightly. labels Oct 28, 2021
@luqmana
Copy link
Member

luqmana commented Oct 28, 2021

cargo-bisect-rustc points at 308dffd (#89179). That being a rollup, #89041 looks the most relevant.

@apiraino
Copy link
Contributor

Assigning priority as discussed in the Zulip thread of the Prioritization Working Group.

@rustbot label -I-prioritize +P-high

@rustbot rustbot added P-high High priority and removed I-prioritize Issue: Indicates that prioritization has been requested for this issue. labels Oct 28, 2021
@apiraino apiraino added the T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. label Oct 28, 2021
@pnkfelix pnkfelix self-assigned this Oct 28, 2021
@apiraino apiraino removed the E-needs-bisection Call for participation: This issue needs bisection: https://github.com/rust-lang/cargo-bisect-rustc label Nov 4, 2021
@cuviper
Copy link
Member

cuviper commented Nov 19, 2021

I think this may be a duplicate of #90301. Please test on a current nightly, and when 1.57.0-beta.4 goes out it should have the backport via #90938.

It's still possible that #89041 introduced its own problem, but maybe it just exacerbated the LLVM issue.

@pnkfelix
Copy link
Member

It looks to me like the problem still reoccurs on the current nightly:

% rustup override set nightly-2021-11-22
info: using existing install for 'nightly-2021-11-22-x86_64-unknown-linux-gnu'
info: override toolchain for '/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro' set to 'nightly-2021-11-22-x86_64-unknown-linux-gnu'

  nightly-2021-11-22-x86_64-unknown-linux-gnu unchanged - rustc 1.58.0-nightly (65f3f8b22 2021-11-21)

% rustup show active-toolchain
nightly-2021-11-22-x86_64-unknown-linux-gnu (directory override for '/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro')
% rustc --version
rustc 1.58.0-nightly (65f3f8b22 2021-11-21)
% cargo clean
% cargo build --release --verbose
   Compiling nodwarfrepro v0.1.0 (/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro)
     Running `rustc --crate-name nodwarfrepro --edition=2018 src/lib.rs --error-format=json --json=diagnostic-rendered-ansi --crate-type lib --emit=dep-info,metadata,link -C opt-level=3 -C linker-plugin-lto -C debuginfo=2 -C metadata=f2ccefc4246296b8 -C extra-filename=-f2ccefc4246296b8 --out-dir /media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro/target/release/deps -L dependency=/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro/target/release/deps`
     Running `rustc --crate-name nodwarfrepro --edition=2018 src/main.rs --error-format=json --json=diagnostic-rendered-ansi --crate-type bin --emit=dep-info,link -C opt-level=3 -C lto -C debuginfo=2 -C metadata=f4b2592c51b39d90 -C extra-filename=-f4b2592c51b39d90 --out-dir /media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro/target/release/deps -L dependency=/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro/target/release/deps --extern nodwarfrepro=/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro/target/release/deps/libnodwarfrepro-f2ccefc4246296b8.rlib`
    Finished release [optimized + debuginfo] target(s) in 2.55s
% dwarfdump target/release/nodwarfrepro | grep THE_STATIC
% 
For completeness, here is a transcript showing the behavior that is expected, from nightly-2021-09-22
% rustup override set nightly-2021-09-22
info: using existing install for 'nightly-2021-09-22-x86_64-unknown-linux-gnu'
info: override toolchain for '/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro' set to 'nightly-2021-09-22-x86_64-unknown-linux-gnu'

  nightly-2021-09-22-x86_64-unknown-linux-gnu unchanged - rustc 1.57.0-nightly (ac2d9fc50 2021-09-21)

% rustup show active-toolchain
nightly-2021-09-22-x86_64-unknown-linux-gnu (directory override for '/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro')
% rustc --version
rustc 1.57.0-nightly (ac2d9fc50 2021-09-21)
% cargo clean
% cargo build --release --verbose
   Compiling nodwarfrepro v0.1.0 (/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro)
     Running `rustc --crate-name nodwarfrepro --edition=2018 src/lib.rs --error-format=json --json=diagnostic-rendered-ansi --crate-type lib --emit=dep-info,metadata,link -C opt-level=3 -C linker-plugin-lto -C debuginfo=2 -C metadata=f2ccefc4246296b8 -C extra-filename=-f2ccefc4246296b8 --out-dir /media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro/target/release/deps -L dependency=/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro/target/release/deps`
     Running `rustc --crate-name nodwarfrepro --edition=2018 src/main.rs --error-format=json --json=diagnostic-rendered-ansi --crate-type bin --emit=dep-info,link -C opt-level=3 -C lto -C debuginfo=2 -C metadata=f4b2592c51b39d90 -C extra-filename=-f4b2592c51b39d90 --out-dir /media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro/target/release/deps -L dependency=/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro/target/release/deps --extern nodwarfrepro=/media/pnkfelix/Rust/rust-90357/objdir-dbgopt/rustc-dwarf-regression-repro/target/release/deps/libnodwarfrepro-f2ccefc4246296b8.rlib`
    Finished release [optimized + debuginfo] target(s) in 2.78s
% dwarfdump target/release/nodwarfrepro | grep THE_STATIC
                        DW_AT_name                  THE_STATIC
                        DW_AT_linkage_name          _ZN12nodwarfrepro10THE_STATIC17he3077d0fbb05bed3E
global die-in-sect 0x00000089, cu-in-sect 0x00000071, die-in-cu 0x00000023, cu-header-in-sect 0x00000066 'THE_STATIC'
name at offset 0x000000d1, length   10 is 'THE_STATIC'
name at offset 0x00000102, length   49 is '_ZN12nodwarfrepro10THE_STATIC17he3077d0fbb05bed3E'
% 

@pnkfelix pnkfelix added regression-from-stable-to-stable Performance or correctness regression from one stable version to another. and removed regression-from-stable-to-beta Performance or correctness regression from stable to beta. labels Dec 2, 2021
Urhengulas pushed a commit to knurling-rs/app-template that referenced this issue Jan 5, 2022
@cbiffle
Copy link
Contributor Author

cbiffle commented Apr 4, 2022

This appears to still be reproducing as of rustc 1.61.0-nightly (9c06e1ba4 2022-03-29); we're stuck on 2021-09-22 because basically anything newer breaks our debugger.

@cbiffle
Copy link
Contributor Author

cbiffle commented Apr 4, 2022

So, after working out how to build a toolchain that includes our cross target and rust-lld and whatnot locally, I've tested today's mainline d5139f4 with a revert for PR #89041, commit d5de680 -- which @luqmana had previously identified as the likely culprit.

I can confirm that DWARF output for our statics on thumbv7em and kin is restored after that commit is reverted. Shall I prepare a PR?

cbiffle added a commit to oxidecomputer/rust that referenced this issue Apr 5, 2022
- Re-enabled basic-types-globals which has been disabled since 2018
- Updated its now-rotted assertions about GDB's output to pass
- Rewrote header comment describing some previous state of GDB behavior
  that didn't match either the checked-in assertions _or_ the current
  behavior, and so I assume has just been wrong for some time.
- Copy-pasta'd the test into a version that uses LTO to check for
  regression on rust-lang#90357, because I don't see a way to matrix the same
  test into several build configurations.
cbiffle added a commit to oxidecomputer/rust that referenced this issue Apr 5, 2022
These tests were disabled four years ago; it appears that it was
intended to be temporary.

In that time their assertions have rotted. I've gone through and fixed
the assertions where necessary and rehabilitated them. In several cases
I've had to leave lldb disabled, because the lldb output has changed in
ways that I don't understand. I figured gdb coverage was better than no
coverage.

It's worth noting that these tests being enabled would probably have
caught the regression observed in rust-lang#90357.
fee1-dead added a commit to fee1-dead-contrib/rust that referenced this issue Apr 15, 2022
… r=pnkfelix

Revert "Work around invalid DWARF bugs for fat LTO"

Since September, the toolchain has not been generating reliable DWARF
information for static variables when LTO is on. This has affected
projects in the embedded space where the use of LTO is typical. In our
case, it has kept us from bumping past the 2021-09-22 nightly toolchain
lest our debugger break. This has been a pretty dramatic regression for
people using debuggers and static variables. See rust-lang#90357 for more info
and a repro case.

This commit is a mechanical revert of
d5de680 from PR rust-lang#89041, which caused
the issue. (Note on that PR that the commit's author has requested it be
reverted.)

I have locally verified that this fixes rust-lang#90357 by restoring the
functionality of both the repro case I posted on that bug, and debugger
behavior on real programs. There do not appear to be test cases for this
in the toolchain; if I've missed them, point me at 'em and I'll update
them.
RalfJung added a commit to RalfJung/rust that referenced this issue Apr 15, 2022
… r=pnkfelix

Revert "Work around invalid DWARF bugs for fat LTO"

Since September, the toolchain has not been generating reliable DWARF
information for static variables when LTO is on. This has affected
projects in the embedded space where the use of LTO is typical. In our
case, it has kept us from bumping past the 2021-09-22 nightly toolchain
lest our debugger break. This has been a pretty dramatic regression for
people using debuggers and static variables. See rust-lang#90357 for more info
and a repro case.

This commit is a mechanical revert of
d5de680 from PR rust-lang#89041, which caused
the issue. (Note on that PR that the commit's author has requested it be
reverted.)

I have locally verified that this fixes rust-lang#90357 by restoring the
functionality of both the repro case I posted on that bug, and debugger
behavior on real programs. There do not appear to be test cases for this
in the toolchain; if I've missed them, point me at 'em and I'll update
them.
RalfJung added a commit to RalfJung/rust that referenced this issue Apr 15, 2022
… r=pnkfelix

Revert "Work around invalid DWARF bugs for fat LTO"

Since September, the toolchain has not been generating reliable DWARF
information for static variables when LTO is on. This has affected
projects in the embedded space where the use of LTO is typical. In our
case, it has kept us from bumping past the 2021-09-22 nightly toolchain
lest our debugger break. This has been a pretty dramatic regression for
people using debuggers and static variables. See rust-lang#90357 for more info
and a repro case.

This commit is a mechanical revert of
d5de680 from PR rust-lang#89041, which caused
the issue. (Note on that PR that the commit's author has requested it be
reverted.)

I have locally verified that this fixes rust-lang#90357 by restoring the
functionality of both the repro case I posted on that bug, and debugger
behavior on real programs. There do not appear to be test cases for this
in the toolchain; if I've missed them, point me at 'em and I'll update
them.
RalfJung added a commit to RalfJung/rust that referenced this issue Apr 15, 2022
… r=pnkfelix

Revert "Work around invalid DWARF bugs for fat LTO"

Since September, the toolchain has not been generating reliable DWARF
information for static variables when LTO is on. This has affected
projects in the embedded space where the use of LTO is typical. In our
case, it has kept us from bumping past the 2021-09-22 nightly toolchain
lest our debugger break. This has been a pretty dramatic regression for
people using debuggers and static variables. See rust-lang#90357 for more info
and a repro case.

This commit is a mechanical revert of
d5de680 from PR rust-lang#89041, which caused
the issue. (Note on that PR that the commit's author has requested it be
reverted.)

I have locally verified that this fixes rust-lang#90357 by restoring the
functionality of both the repro case I posted on that bug, and debugger
behavior on real programs. There do not appear to be test cases for this
in the toolchain; if I've missed them, point me at 'em and I'll update
them.
Dylan-DPC added a commit to Dylan-DPC/rust that referenced this issue Apr 15, 2022
… r=pnkfelix

Revert "Work around invalid DWARF bugs for fat LTO"

Since September, the toolchain has not been generating reliable DWARF
information for static variables when LTO is on. This has affected
projects in the embedded space where the use of LTO is typical. In our
case, it has kept us from bumping past the 2021-09-22 nightly toolchain
lest our debugger break. This has been a pretty dramatic regression for
people using debuggers and static variables. See rust-lang#90357 for more info
and a repro case.

This commit is a mechanical revert of
d5de680 from PR rust-lang#89041, which caused
the issue. (Note on that PR that the commit's author has requested it be
reverted.)

I have locally verified that this fixes rust-lang#90357 by restoring the
functionality of both the repro case I posted on that bug, and debugger
behavior on real programs. There do not appear to be test cases for this
in the toolchain; if I've missed them, point me at 'em and I'll update
them.
bors added a commit to rust-lang-ci/rust that referenced this issue Jul 16, 2022
…=pnkfelix

Revert "Work around invalid DWARF bugs for fat LTO"

Since September, the toolchain has not been generating reliable DWARF
information for static variables when LTO is on. This has affected
projects in the embedded space where the use of LTO is typical. In our
case, it has kept us from bumping past the 2021-09-22 nightly toolchain
lest our debugger break. This has been a pretty dramatic regression for
people using debuggers and static variables. See rust-lang#90357 for more info
and a repro case.

This commit is a mechanical revert of
d5de680 from PR rust-lang#89041, which caused
the issue. (Note on that PR that the commit's author has requested it be
reverted.)

I have locally verified that this fixes rust-lang#90357 by restoring the
functionality of both the repro case I posted on that bug, and debugger
behavior on real programs. There do not appear to be test cases for this
in the toolchain; if I've missed them, point me at 'em and I'll update
them.
@bors bors closed this as completed in 98190b7 Jul 16, 2022
@jieyouxu jieyouxu added A-reproducibility Area: Reproducible / Deterministic builds A-LLVM Area: Code generation parts specific to LLVM. Both correctness bugs and optimization-related issues. labels Aug 14, 2024
@jieyouxu jieyouxu added the A-lto Area: Link Time Optimization label Aug 14, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
A-debuginfo Area: Debugging information in compiled programs (DWARF, PDB, etc.) A-LLVM Area: Code generation parts specific to LLVM. Both correctness bugs and optimization-related issues. A-lto Area: Link Time Optimization A-reproducibility Area: Reproducible / Deterministic builds C-bug Category: This is a bug. P-high High priority regression-from-stable-to-stable Performance or correctness regression from one stable version to another. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue.
Projects
None yet
Development

Successfully merging a pull request may close this issue.

9 participants