perf:cache FFI safety results in improper_ctypes lint by xonx4l · Pull Request #163348 · rust-lang/rust

xonx4l · 2026-09-25T18:17:15Z

This PR cache FFI safety results in improper_ctypes lint .

ImproperCTypesVisitor's cache field is recreated empty at every call site, so a struct type gets fully re-walked from scratch each time instead of once .

The implementation here moves the cache onto ImproperCtypesLint, so it remembers types we already checked and found FFI-safe, so we don't check the same type again every time it shows up in another function. We only remember "safe" types, never "unsafe" ones. That way we never need to keep the actual type around, just a fingerprint of it.Thus it persists across a whole module worth of foreign items.

rustbot · 2026-09-25T18:17:20Z

r? @nnethercote

rustbot has assigned @nnethercote.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

Owners of files modified in this PR: compiler
compiler expanded to 77 candidates
Random selection from 19 candidates

nnethercote · 2026-09-26T04:53:25Z

What motivated this change? Do you have any measurements showing that it improves compile times?

xonx4l · 2026-09-28T05:26:48Z

The motivation for the change is that the cache is recreated empty at every call site, so a struct shared across many extern functions gets fully re-checked from scratch each time instead of once.

For measurement I profiled a test file containing some extern "C" functions sharing a few structs with cachegrind :).

Before: 2,000,537,577 instructions count , After: 1,753,411,352 instructions count

Which gives approx a total of 12% of instructions count reduction. Thank you!

nnethercote · 2026-10-02T02:57:57Z

+        let ffi_res = visitor.check_type(state, ty);
+        if matches!(ffi_res, FfiResult::FfiSafe) {
+            self.known_safe.borrow_mut().insert(key);
+        }


It's unfortunate to have this contains + "do stuff" + insert pattern repeated three times. Can you factor it out into a separate method? It might need a closure argument for the "do stuff" part.

@rustbot author

View changes since the review

Ok, will make it a separate method . Thank you!

rustbot · 2026-10-02T02:58:00Z

Reminder, once the PR becomes ready for a review, use @rustbot ready.

xonx4l · 2026-10-05T19:11:31Z

@rustbot ready

nnethercote · 2026-10-06T04:28:35Z

@bors r+ rollup

rust-bors · 2026-10-06T04:28:39Z

📌 Commit 54a6d2e has been approved by nnethercote

It is now in the queue for this repository.

… r=nnethercote perf:cache FFI safety results in improper_ctypes lint This PR cache FFI safety results in improper_ctypes lint . ImproperCTypesVisitor's cache field is recreated empty at every call site, so a struct type gets fully re-walked from scratch each time instead of once . The implementation here moves the cache onto ImproperCtypesLint, so it remembers types we already checked and found FFI-safe, so we don't check the same type again every time it shows up in another function. We only remember "safe" types, never "unsafe" ones. That way we never need to keep the actual type around, just a fingerprint of it.Thus it persists across a whole module worth of foreign items.

Rollup of 3 pull requests Successful merges: - #163809 (Adjust the adjustment to wasi TLS to no longer adjust) - #163348 (perf:cache FFI safety results in improper_ctypes lint) - #163772 (comptime fn error: suggest wrapping in const block)

rust-bors · 2026-10-06T04:57:44Z

⌛ Testing commit 54a6d2e with merge b1bfbc8...

Workflow: https://github.com/rust-lang/rust/actions/runs/37416185729

perf:cache FFI safety results in improper_ctypes lint This PR cache FFI safety results in improper_ctypes lint . ImproperCTypesVisitor's cache field is recreated empty at every call site, so a struct type gets fully re-walked from scratch each time instead of once . The implementation here moves the cache onto ImproperCtypesLint, so it remembers types we already checked and found FFI-safe, so we don't check the same type again every time it shows up in another function. We only remember "safe" types, never "unsafe" ones. That way we never need to keep the actual type around, just a fingerprint of it.Thus it persists across a whole module worth of foreign items.

… r=nnethercote perf:cache FFI safety results in improper_ctypes lint This PR cache FFI safety results in improper_ctypes lint . ImproperCTypesVisitor's cache field is recreated empty at every call site, so a struct type gets fully re-walked from scratch each time instead of once . The implementation here moves the cache onto ImproperCtypesLint, so it remembers types we already checked and found FFI-safe, so we don't check the same type again every time it shows up in another function. We only remember "safe" types, never "unsafe" ones. That way we never need to keep the actual type around, just a fingerprint of it.Thus it persists across a whole module worth of foreign items.

jhpratt · 2026-10-06T05:04:09Z

@bors yield once more

rust-bors · 2026-10-06T05:04:14Z

Auto build was cancelled. Cancelled workflows:

https://github.com/rust-lang/rust/actions/runs/37416185729

The next pull request likely to be tested is #163859.

rust-bors · 2026-10-06T09:07:25Z

Auto build was cancelled. It was not possible to cancel some workflows.

The next pull request likely to be tested is #163831.

… r=nnethercote perf:cache FFI safety results in improper_ctypes lint This PR cache FFI safety results in improper_ctypes lint . ImproperCTypesVisitor's cache field is recreated empty at every call site, so a struct type gets fully re-walked from scratch each time instead of once . The implementation here moves the cache onto ImproperCtypesLint, so it remembers types we already checked and found FFI-safe, so we don't check the same type again every time it shows up in another function. We only remember "safe" types, never "unsafe" ones. That way we never need to keep the actual type around, just a fingerprint of it.Thus it persists across a whole module worth of foreign items.

…uwer Rollup of 13 pull requests Successful merges: - #163090 (Run LLDB debuginfo tests on `x86_64-pc-windows-msvc` in CI) - #163863 (use the type name instead of `Self` in constructor paths again) - #163864 (Remove CLAUDE.md) - #163877 (Add regression test for duplicated rustdoc search results between std and core) - #162000 (peel_transparent_wrappers only works on non-1ZST) - #163257 (abby DSL: sanity checks on forall where clauses) - #163348 (perf:cache FFI safety results in improper_ctypes lint) - #163768 (Move more `rustdoc-html` tests in the right location) - #163789 (cg_llvm: Avoid `as_c_char_ptr` in several places) - #163807 (Add `has_reliable_f16b` for Arm) - #163826 (Update books) - #163849 (ci: update to PowerShell 7.6.6) - #163870 (Add if-installed:auto:spellcheck to pre-push script)

… r=nnethercote perf:cache FFI safety results in improper_ctypes lint This PR cache FFI safety results in improper_ctypes lint . ImproperCTypesVisitor's cache field is recreated empty at every call site, so a struct type gets fully re-walked from scratch each time instead of once . The implementation here moves the cache onto ImproperCtypesLint, so it remembers types we already checked and found FFI-safe, so we don't check the same type again every time it shows up in another function. We only remember "safe" types, never "unsafe" ones. That way we never need to keep the actual type around, just a fingerprint of it.Thus it persists across a whole module worth of foreign items.

…uwer Rollup of 17 pull requests Successful merges: - #163090 (Run LLDB debuginfo tests on `x86_64-pc-windows-msvc` in CI) - #163794 (Use `target_family = "wasm"` instead of `target_arch = "wasm32"`) - #163863 (use the type name instead of `Self` in constructor paths again) - #163864 (Remove CLAUDE.md) - #163877 (Add regression test for duplicated rustdoc search results between std and core) - #162000 (peel_transparent_wrappers only works on non-1ZST) - #163257 (abby DSL: sanity checks on forall where clauses) - #163348 (perf:cache FFI safety results in improper_ctypes lint) - #163768 (Move more `rustdoc-html` tests in the right location) - #163778 (check is_indirect_first_projection when replacing in RefProp) - #163789 (cg_llvm: Avoid `as_c_char_ptr` in several places) - #163807 (Add `has_reliable_f16b` for Arm) - #163826 (Update books) - #163849 (ci: update to PowerShell 7.6.6) - #163870 (Add if-installed:auto:spellcheck to pre-push script) - #163874 (explicitly check for fresh vars in canonicalize) - #163880 (Don't add pkgs.rustc to PATH in nix dev shell)

… r=nnethercote perf:cache FFI safety results in improper_ctypes lint This PR cache FFI safety results in improper_ctypes lint . ImproperCTypesVisitor's cache field is recreated empty at every call site, so a struct type gets fully re-walked from scratch each time instead of once . The implementation here moves the cache onto ImproperCtypesLint, so it remembers types we already checked and found FFI-safe, so we don't check the same type again every time it shows up in another function. We only remember "safe" types, never "unsafe" ones. That way we never need to keep the actual type around, just a fingerprint of it.Thus it persists across a whole module worth of foreign items.

…uwer Rollup of 25 pull requests Successful merges: - #163090 (Run LLDB debuginfo tests on `x86_64-pc-windows-msvc` in CI) - #163806 (even more `tests/crashes` migration for `-Znext-solver`) - #162156 (add IBM f128 type) - #163508 (Document the `rustc_on_unimplemented` attribute.) - #163794 (Use `target_family = "wasm"` instead of `target_arch = "wasm32"`) - #163863 (use the type name instead of `Self` in constructor paths again) - #163864 (Remove CLAUDE.md) - #163877 (Add regression test for duplicated rustdoc search results between std and core) - #163916 (Shrink `PartialRes` and its alignment) - #149753 (On name resolution error in parameter list, suggest possible `const` typo and avoid unnecessary second error) - #152557 (Oneshot `is_ready`) - #157273 (Stabilize `optimize` attribute) - #162000 (peel_transparent_wrappers only works on non-1ZST) - #163257 (abby DSL: sanity checks on forall where clauses) - #163343 (`unit` and `[1 x unit]` layouts are incompatible on big-endian targets) - #163348 (perf:cache FFI safety results in improper_ctypes lint) - #163768 (Move more `rustdoc-html` tests in the right location) - #163778 (check is_indirect_first_projection when replacing in RefProp) - #163789 (cg_llvm: Avoid `as_c_char_ptr` in several places) - #163807 (Add `has_reliable_f16b` for Arm) - #163809 (Switch TLS implementation for wasi and bump SDK version to 34) - #163826 (Update books) - #163849 (ci: update to PowerShell 7.6.6) - #163869 (INSTALL.md: use UCRT64 instead of MINGW64 in MSYS2 section) - #163870 (Add if-installed:auto:spellcheck to pre-push script)

… r=nnethercote perf:cache FFI safety results in improper_ctypes lint This PR cache FFI safety results in improper_ctypes lint . ImproperCTypesVisitor's cache field is recreated empty at every call site, so a struct type gets fully re-walked from scratch each time instead of once . The implementation here moves the cache onto ImproperCtypesLint, so it remembers types we already checked and found FFI-safe, so we don't check the same type again every time it shows up in another function. We only remember "safe" types, never "unsafe" ones. That way we never need to keep the actual type around, just a fingerprint of it.Thus it persists across a whole module worth of foreign items.

…uwer Rollup of 24 pull requests Successful merges: - #163090 (Run LLDB debuginfo tests on `x86_64-pc-windows-msvc` in CI) - #163806 (even more `tests/crashes` migration for `-Znext-solver`) - #163508 (Document the `rustc_on_unimplemented` attribute.) - #163794 (Use `target_family = "wasm"` instead of `target_arch = "wasm32"`) - #163863 (use the type name instead of `Self` in constructor paths again) - #163864 (Remove CLAUDE.md) - #163877 (Add regression test for duplicated rustdoc search results between std and core) - #163916 (Shrink `PartialRes` and its alignment) - #149753 (On name resolution error in parameter list, suggest possible `const` typo and avoid unnecessary second error) - #152557 (Oneshot `is_ready`) - #157273 (Stabilize `optimize` attribute) - #162000 (peel_transparent_wrappers only works on non-1ZST) - #163257 (abby DSL: sanity checks on forall where clauses) - #163343 (`unit` and `[1 x unit]` layouts are incompatible on big-endian targets) - #163348 (perf:cache FFI safety results in improper_ctypes lint) - #163768 (Move more `rustdoc-html` tests in the right location) - #163778 (check is_indirect_first_projection when replacing in RefProp) - #163789 (cg_llvm: Avoid `as_c_char_ptr` in several places) - #163807 (Add `has_reliable_f16b` for Arm) - #163809 (Switch TLS implementation for wasi and bump SDK version to 34) - #163826 (Update books) - #163849 (ci: update to PowerShell 7.6.6) - #163869 (INSTALL.md: use UCRT64 instead of MINGW64 in MSYS2 section) - #163870 (Add if-installed:auto:spellcheck to pre-push script)

…uwer Rollup of 23 pull requests Successful merges: - #163090 (Run LLDB debuginfo tests on `x86_64-pc-windows-msvc` in CI) - #163806 (even more `tests/crashes` migration for `-Znext-solver`) - #163508 (Document the `rustc_on_unimplemented` attribute.) - #163794 (Use `target_family = "wasm"` instead of `target_arch = "wasm32"`) - #163863 (use the type name instead of `Self` in constructor paths again) - #163864 (Remove CLAUDE.md) - #163877 (Add regression test for duplicated rustdoc search results between std and core) - #163916 (Shrink `PartialRes` and its alignment) - #149753 (On name resolution error in parameter list, suggest possible `const` typo and avoid unnecessary second error) - #152557 (Oneshot `is_ready`) - #157273 (Stabilize `optimize` attribute) - #162000 (peel_transparent_wrappers only works on non-1ZST) - #163343 (`unit` and `[1 x unit]` layouts are incompatible on big-endian targets) - #163348 (perf:cache FFI safety results in improper_ctypes lint) - #163768 (Move more `rustdoc-html` tests in the right location) - #163778 (check is_indirect_first_projection when replacing in RefProp) - #163789 (cg_llvm: Avoid `as_c_char_ptr` in several places) - #163807 (Add `has_reliable_f16b` for Arm) - #163809 (Switch TLS implementation for wasi and bump SDK version to 34) - #163826 (Update books) - #163849 (ci: update to PowerShell 7.6.6) - #163869 (INSTALL.md: use UCRT64 instead of MINGW64 in MSYS2 section) - #163870 (Add if-installed:auto:spellcheck to pre-push script)

Rollup merge of #163348 - xonx4l:improper_ctypes-lint-cache, r=nnethercote perf:cache FFI safety results in improper_ctypes lint This PR cache FFI safety results in improper_ctypes lint . ImproperCTypesVisitor's cache field is recreated empty at every call site, so a struct type gets fully re-walked from scratch each time instead of once . The implementation here moves the cache onto ImproperCtypesLint, so it remembers types we already checked and found FFI-safe, so we don't check the same type again every time it shows up in another function. We only remember "safe" types, never "unsafe" ones. That way we never need to keep the actual type around, just a fingerprint of it.Thus it persists across a whole module worth of foreign items.

rust-timer · 2026-10-08T06:05:22Z

Note

This PR was benchmarked as part of triage of its containing rollup: triage URL.

Finished benchmarking commit (e5ce1f7): comparison URL.

Overall result: ✅ improvements - no action needed

@rustbot label: -perf-regression

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.1%	[-0.3%, -0.1%]	4
Improvements ✅ (secondary)	-0.3%	[-0.7%, -0.1%]	21
All ❌✅ (primary)	-0.1%	[-0.3%, -0.1%]	4

Max RSS (memory usage)

This perf run didn't have relevant results for this metric.

Cycles

This perf run didn't have relevant results for this metric.

Binary size

This perf run didn't have relevant results for this metric.

Artifact size: 408.59 MiB -> 408.61 MiB (0.00%)

JonathanBrouwer · 2026-10-08T16:23:09Z

This is noise, all PRs in the rollup have the same results

cache results in improper_ctypes lint

5e9dca9

rustbot assigned nnethercote Sep 25, 2026

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Sep 25, 2026

This comment has been minimized.

Sign in to view

nnethercote added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Sep 27, 2026

fix build errors

fa3bfde

nnethercote reviewed Oct 2, 2026

View reviewed changes

factor out as check_ffi_type method

54a6d2e

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Oct 5, 2026

rust-bors Bot added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Oct 6, 2026

jhpratt mentioned this pull request Oct 6, 2026

Rollup of 3 pull requests #163858

Closed

jhpratt mentioned this pull request Oct 6, 2026

Rollup of 2 pull requests #163859

Closed

JonathanBrouwer mentioned this pull request Oct 6, 2026

Rollup of 6 pull requests #163873

Closed

JonathanBrouwer mentioned this pull request Oct 6, 2026

Rollup of 8 pull requests #163876

Closed

JonathanBrouwer mentioned this pull request Oct 6, 2026

Rollup of 13 pull requests #163879

Closed

JonathanBrouwer mentioned this pull request Oct 6, 2026

Rollup of 17 pull requests #163884

Closed

JonathanBrouwer mentioned this pull request Oct 7, 2026

Rollup of 25 pull requests #163919

Closed

JonathanBrouwer mentioned this pull request Oct 7, 2026

Rollup of 24 pull requests #163934

Closed

JonathanBrouwer mentioned this pull request Oct 7, 2026

Rollup of 23 pull requests #163957

Merged

rust-bors Bot merged commit dfa9cf1 into rust-lang:main Oct 8, 2026
14 of 15 checks passed

rustbot added this to the 1.101.0 milestone Oct 8, 2026

github-actions Bot mentioned this pull request Oct 9, 2026

Toolchain upgrade to nightly-2026-10-09 flux-rs/flux#1811

Draft

Conversation

xonx4l commented Sep 25, 2026

Uh oh!

rustbot commented Sep 25, 2026

Uh oh!

This comment has been minimized.

nnethercote commented Sep 26, 2026

Uh oh!

xonx4l commented Sep 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

nnethercote Oct 2, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xonx4l Oct 2, 2026

Choose a reason for hiding this comment

Uh oh!

rustbot commented Oct 2, 2026

Uh oh!

xonx4l commented Oct 5, 2026

Uh oh!

nnethercote commented Oct 6, 2026

Uh oh!

rust-bors Bot commented Oct 6, 2026

Uh oh!

rust-bors Bot commented Oct 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jhpratt commented Oct 6, 2026

Uh oh!

rust-bors Bot commented Oct 6, 2026

Uh oh!

rust-bors Bot commented Oct 6, 2026

Uh oh!

Uh oh!

rust-timer commented Oct 8, 2026

Overall result: ✅ improvements - no action needed

Instruction count

Max RSS (memory usage)

Cycles

Binary size

Uh oh!

JonathanBrouwer commented Oct 8, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

xonx4l commented Sep 28, 2026 •

edited

Loading

nnethercote Oct 2, 2026 •

edited by rustbot

Loading

rust-bors Bot commented Oct 6, 2026 •

edited

Loading