Handle multiple action records in EH personality function - #160923
Merged
Conversation
rustbot
added
S-waiting-on-author
This comment has been minimized.
This comment has been minimized.
Contributor
|
Thanks for fixing this! I did the test on my side and it seems to work. and with this PR, I get the correct behavior in my reproducer. |
nbdd0121
force-pushed
the
eh_personality
branch
from
August 11, 2026 17:43
21dbc6e to
6663fc5
Compare
This comment has been minimized.
This comment has been minimized.
nbdd0121
force-pushed
the
eh_personality
branch
from
August 11, 2026 18:03
6663fc5 to
c25db43
Compare
Member
Author
|
r? bjorn3 perhaps? |
nbdd0121
marked this pull request as ready for review
August 11, 2026 19:11
rustbot
added
S-waiting-on-review
nbdd0121
force-pushed
the
eh_personality
branch
from
August 11, 2026 19:12
c25db43 to
689d39d
Compare
bjorn3
reviewed
Aug 27, 2026
| // Note that even for the case of "cleanup" + "filter", decoding them as "catch" is | ||
| // fine: "filter" behaves identically to "catch" except for forced unwind; in case of | ||
| // forced unwind, hitting a "cleanup" landing pad is UB as it indicates that we're | ||
| // unwinding past a non-POF Rust frame. |
Member
There was a problem hiding this comment.
What if there are two cleanup actions for whatever reason? Should this explicitly check that at least one action is a filter or catch if there are multiple ones and abort otherwise?
Member
Author
There was a problem hiding this comment.
That'll be a compiler bug? I don't think we should defend against something that won't happen at a cost of code size.
Also, nothing will be catastrophically wrong even if we catch a cleanup frame, we will just be skipping phase 1 of unwind and perform phase 2 unwind for further stack frames, similar to forced unwind.
Member
There was a problem hiding this comment.
Wouldn't the unwinder misbehave if you _Unwind_Resume out of a frame that you said is a catch?
Member
Author
There was a problem hiding this comment.
It might unwind more frames, but nothing will be terribly wrong. Anyhow, neither GCC nor LLVM will emit duplicate action record entry, so I don't think this is worth handling.
bjorn3
added
S-waiting-on-author
bjorn3
reviewed
Aug 28, 2026
nbdd0121
force-pushed
the
eh_personality
branch
from
August 29, 2026 22:49
689d39d to
bf56fa4
Compare
Member
|
@bors r+ |
Contributor
rust-bors
Bot
added
S-waiting-on-bors
JonathanBrouwer
added a commit
to JonathanBrouwer/rust
that referenced
this pull request
Aug 30, 2026
Handle multiple action records in EH personality function LSDA encodes actions are a linked list, we currently only decode the first one. Action records are used for catching specific exception types and for exception specifications, and none of these are present in Rust. However we can still have of multiple of them being present due to LLVM inlining. When this happens, "Catch" is the correct action to execute. We haven't had issues with this because LLVM orders cleanup record last; however GCC doesn't use the same order, so we do need to explicitly handle this case to support both LLVM and GCC codegen. See https://rust-lang.zulipchat.com/#narrow/channel/182449-t-compiler.2Fhelp/topic/Possible.20bug.20in.20unwind.20function.20in.20std/with/615883393
JonathanBrouwer
added a commit
to JonathanBrouwer/rust
that referenced
this pull request
Aug 30, 2026
Handle multiple action records in EH personality function LSDA encodes actions are a linked list, we currently only decode the first one. Action records are used for catching specific exception types and for exception specifications, and none of these are present in Rust. However we can still have of multiple of them being present due to LLVM inlining. When this happens, "Catch" is the correct action to execute. We haven't had issues with this because LLVM orders cleanup record last; however GCC doesn't use the same order, so we do need to explicitly handle this case to support both LLVM and GCC codegen. See https://rust-lang.zulipchat.com/#narrow/channel/182449-t-compiler.2Fhelp/topic/Possible.20bug.20in.20unwind.20function.20in.20std/with/615883393
rust-bors Bot
pushed a commit
that referenced
this pull request
Aug 30, 2026
…uwer Rollup of 8 pull requests Successful merges: - #137720 (support `#[target_feature(enable = ...)]` on `#[naked]` functions) - #160923 (Handle multiple action records in EH personality function) - #161788 (Check to ensure we're running against the correct LLVM version) - #161644 (bootstrap: Flatten and rename `compute_src_directory_via_git`) - #161702 (Use `drop_guard` in some places in {core,alloc,std}) - #161931 (Revert "Add rustc_test_entrypoint_marker") - #162015 (touch up "get attribute" docs.) - #162019 (remove a couple of redundant clones, thanks clippy)
JonathanBrouwer
added a commit
to JonathanBrouwer/rust
that referenced
this pull request
Aug 30, 2026
Handle multiple action records in EH personality function LSDA encodes actions are a linked list, we currently only decode the first one. Action records are used for catching specific exception types and for exception specifications, and none of these are present in Rust. However we can still have of multiple of them being present due to LLVM inlining. When this happens, "Catch" is the correct action to execute. We haven't had issues with this because LLVM orders cleanup record last; however GCC doesn't use the same order, so we do need to explicitly handle this case to support both LLVM and GCC codegen. See https://rust-lang.zulipchat.com/#narrow/channel/182449-t-compiler.2Fhelp/topic/Possible.20bug.20in.20unwind.20function.20in.20std/with/615883393
JonathanBrouwer
added a commit
to JonathanBrouwer/rust
that referenced
this pull request
Aug 30, 2026
Handle multiple action records in EH personality function LSDA encodes actions are a linked list, we currently only decode the first one. Action records are used for catching specific exception types and for exception specifications, and none of these are present in Rust. However we can still have of multiple of them being present due to LLVM inlining. When this happens, "Catch" is the correct action to execute. We haven't had issues with this because LLVM orders cleanup record last; however GCC doesn't use the same order, so we do need to explicitly handle this case to support both LLVM and GCC codegen. See https://rust-lang.zulipchat.com/#narrow/channel/182449-t-compiler.2Fhelp/topic/Possible.20bug.20in.20unwind.20function.20in.20std/with/615883393
This was referenced Aug 30, 2026
rust-bors Bot
pushed a commit
that referenced
this pull request
Aug 30, 2026
…uwer Rollup of 11 pull requests Successful merges: - #160923 (Handle multiple action records in EH personality function) - #161788 (Check to ensure we're running against the correct LLVM version) - #158303 (Always escape grapheme extenders in `str::escape_debug`) - #161644 (bootstrap: Flatten and rename `compute_src_directory_via_git`) - #161702 (Use `drop_guard` in some places in {core,alloc,std}) - #161713 (Add config for gram editor to .gitignore) - #161792 (Fix and improve diagnostics for lint `rust_2021_prefixes_incompatible_syntax`) - #161931 (Revert "Add rustc_test_entrypoint_marker") - #162015 (touch up "get attribute" docs.) - #162019 (remove a couple of redundant clones, thanks clippy) - #162023 (Remove redundant braces from `NonZero` doctests)
rust-bors Bot
pushed a commit
that referenced
this pull request
Aug 30, 2026
…uwer Rollup of 11 pull requests Successful merges: - #160923 (Handle multiple action records in EH personality function) - #161788 (Check to ensure we're running against the correct LLVM version) - #158303 (Always escape grapheme extenders in `str::escape_debug`) - #161644 (bootstrap: Flatten and rename `compute_src_directory_via_git`) - #161702 (Use `drop_guard` in some places in {core,alloc,std}) - #161713 (Add config for gram editor to .gitignore) - #161792 (Fix and improve diagnostics for lint `rust_2021_prefixes_incompatible_syntax`) - #161931 (Revert "Add rustc_test_entrypoint_marker") - #162015 (touch up "get attribute" docs.) - #162019 (remove a couple of redundant clones, thanks clippy) - #162023 (Remove redundant braces from `NonZero` doctests)
rust-bors Bot
pushed a commit
that referenced
this pull request
Aug 31, 2026
Rollup merge of #160923 - nbdd0121:eh_personality, r=bjorn3 Handle multiple action records in EH personality function LSDA encodes actions are a linked list, we currently only decode the first one. Action records are used for catching specific exception types and for exception specifications, and none of these are present in Rust. However we can still have of multiple of them being present due to LLVM inlining. When this happens, "Catch" is the correct action to execute. We haven't had issues with this because LLVM orders cleanup record last; however GCC doesn't use the same order, so we do need to explicitly handle this case to support both LLVM and GCC codegen. See https://rust-lang.zulipchat.com/#narrow/channel/182449-t-compiler.2Fhelp/topic/Possible.20bug.20in.20unwind.20function.20in.20std/with/615883393
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
LSDA encodes actions are a linked list, we currently only decode the first one. Action records are used for catching specific exception types and for exception specifications, and none of these are present in Rust. However we can still have of multiple of them being present due to LLVM inlining. When this happens, "Catch" is the correct action to execute.
We haven't had issues with this because LLVM orders cleanup record last; however GCC doesn't use the same order, so we do need to explicitly handle this case to support both LLVM and GCC codegen.
See https://rust-lang.zulipchat.com/#narrow/channel/182449-t-compiler.2Fhelp/topic/Possible.20bug.20in.20unwind.20function.20in.20std/with/615883393