Do not remove trivial `SwitchInt` with mir-opt-level=0 #139042

compiler-errors · 2025-03-28T03:31:48Z

This PR ensures that we don't prematurely remove trivial SwitchInt terminators which affects both the borrow-checking and runtime semantics (i.e. UB) of the code. Previously the SimplifyCfg optimization was removing SwitchInt terminators when they was "trivial", i.e. when all arms branched to the same basic block, even if that SwitchInt terminator had the side-effect of reading an operand which (for example) may not be initialized or may point to an invalid place in memory.

This behavior is unlike all other optimizations, which are only applied after "analysis" (i.e. borrow-checking) is finished, and which Miri disables to make sure the compiler doesn't silently remove UB.

Fixing this code "breaks" (i.e. unmasks) code that used to borrow-check but no longer does, like:

fn foo() {
    let x;
    let (0 | _) = x;
}

This match expression should perform a read because _ does not shadow the 0 literal pattern, and the compiler should have to read the match scrutinee to compare it to 0. I've checked that this behavior does not actually manifest in practice via a crater run which came back clean: #139042 (comment)

As a side-note, it may be tempting to suggest that this is actually a good thing or that we should preserve this behavior. If we wanted to make this work (i.e. trivially optimize out reads from matches that are redundant like 0 | _), then we should be enabling this behavior after fixing this. However, I think it's kinda unprincipled, and for example other variations of the code don't even work today, e.g.:

fn foo() {
    let x;
    let (0.. | _) = x;
}

rustbot · 2025-03-28T03:31:52Z

r? @saethlin

rustbot has assigned @saethlin.
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

rustbot · 2025-03-28T03:31:54Z

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

The Miri subtree was changed

cc @rust-lang/miri

compiler-errors · 2025-03-28T03:37:16Z

compiler/rustc_mir_transform/src/simplify.rs

-            _ => return false,
+        // Removing a `SwitchInt` terminator may remove reads that result in UB,
+        // so we must not apply this optimization when mir-opt-level = 0.
+        if self.mir_opt_level == 0 {


We could perhaps optimize trivial operands... Like plain locals. Maybe consts?

Well I guess even plain locals could have UB if they're uninitialized. But Rust would never produce that 🤔

SimplifyCfg runs before borrowck, so the following compiles:

fn main() { let bad_ref: &i32; let &(0 | _) = bad_ref; }

oh wonderful, so this is a load bearing optimization then

Ah, not just a bug for Miri then but a bigger problem. Certainly we shouldn't optimize before borrowck.

So instead of checking the mir_opt_level, this should be told which "phase" it is in (or it can maybe even get that info from the MIR body, not sure), and then not do this transformation in the "analysis" phase.

compiler-errors · 2025-03-28T05:39:47Z

well i do wonder if anything relies on this today

@bors try

…chint, r=<try> Do not remove trivial `SwitchInt` with mir-opt-level=0 When mir-opt-level=0, do not optimize out `SwitchInt` terminators that all have the same terminator since it may remove a read which affects miri's ability to detect UB on that operand. cc `@RalfJung` Fixes rust-lang/miri#4237 This affects some tests... I guess I could mark them as `mir-opt-level=1`? Not sure.

bors · 2025-03-28T05:40:59Z

⌛ Trying commit 6bad178 with merge 4dba42b...

compiler-errors · 2025-03-28T05:47:42Z

@rustbot author

RalfJung · 2025-03-28T07:14:39Z

src/tools/miri/tests/fail/read_from_trivial_switch.rs

@@ -0,0 +1,8 @@
+use std::mem::MaybeUninit;


Please add a top-level doc comment explaining what this is testing and referencing the issue.

This comment is still unresolved.

bors · 2025-03-28T07:46:55Z

☀️ Try build successful - checks-actions
Build commit: 4dba42b (4dba42b29bacedb63812cd6f781e31ba4b782a95)

oli-obk · 2025-03-28T08:22:29Z

@craterbot check

craterbot · 2025-03-28T08:22:44Z

👌 Experiment pr-139042 created and queued.
🤖 Automatically detected try build 4dba42b
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

tmiasko · 2025-03-28T08:35:13Z

Pull request doesn't resolve #139042 (comment) yet and mir-opt-level=0 is never used by default. Are we getting ahead of ourselves with crater?

RalfJung · 2025-03-28T08:47:50Z

Yeah I think we do.
@craterbot cancel

The PR first needs to be changed to never do this SwitchInt removal in SimplifyCfg::Initial, SimplifyCfg::PromoteConsts, SimplifyCfg::RemoveFalseEdges, SimplifyCfg::PostAnalysis. And for SimplifyCfg::PreOptimizations it should likely be gated on mir_opt_level. Maintaining that list seems fragile so hopefully there is a better way... maybe it is checking for the phase and the opt_level.

Alternatively, instead of checking the opt level, we could have a separate flag. We already have a similar problem with RemovePlaceMention, which is now gated on the mir_keep_place_mention Z-flag. We could instead have a mir_preserve_ub Z-flag that controls both RemovePlaceMention and SimplifyCfg.

craterbot · 2025-03-28T08:50:37Z

🗑️ Experiment pr-139042 deleted!

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

compiler-errors · 2025-03-30T19:02:25Z

@bors try

bors · 2025-03-30T19:03:36Z

⌛ Trying commit 01ee75d with merge 5fee451...

…chint, r=<try> Do not remove trivial `SwitchInt` with mir-opt-level=0 When mir-opt-level=0, do not optimize out `SwitchInt` terminators that all have the same terminator since it may remove a read which affects miri's ability to detect UB on that operand. cc `@RalfJung` Fixes rust-lang/miri#4237 This affects some tests... I guess I could mark them as `mir-opt-level=1`? Not sure.

bors · 2025-03-30T21:09:56Z

☀️ Try build successful - checks-actions
Build commit: 5fee451 (5fee451ebd94296ea05214e819c311e164b30af6)

compiler-errors · 2025-03-30T23:11:37Z

@craterbot check

craterbot · 2025-03-30T23:11:54Z

👌 Experiment pr-139042 created and queued.
🤖 Automatically detected try build 5fee451
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

craterbot · 2025-03-30T23:13:34Z

🚧 Experiment pr-139042 is now running

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

craterbot · 2025-03-31T21:26:01Z

🎉 Experiment pr-139042 is completed!
📊 5 regressed and 5 fixed (606182 total)
📰 Open the full report.

⚠️ If you notice any spurious failure please add them to the denylist!
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

compiler-errors · 2025-03-31T21:30:40Z

Oh good. Looks like nobody relies on this behavior.

@RalfJung: I took your suggested approach and generalized the flag into -Zmir-preserve-ub and then disabled SwitchInt optimization in the Built phase, since that's the pre-Analysis(Initial) state that we use for borrowck.

I can confirm that causes the example that @tmiasko shared to fail to compile (tho I'll add a test).

I'll write up a better description and then probably FCP this with lang?

compiler/rustc_mir_transform/src/simplify.rs

compiler-errors · 2025-04-07T19:10:04Z

@rustbot label: -T-compiler +T-lang +I-lang-nominated +I-lang-easy-decision

Hi T-lang, this needs an FCP because it's technically breakage. See the above PR description for details; this isn't something we should've ever accepted, IMO.

compiler-errors · 2025-04-07T19:10:57Z

r? @saethlin (or perhaps @RalfJung) or reassign, just choosing a name out of a hat.

rustbot · 2025-04-07T19:10:59Z

Could not assign reviewer from: saethlin.
User(s) saethlin are either the PR author, already assigned, or on vacation. Please use r? to specify someone else to assign.

compiler-errors · 2025-04-07T19:15:41Z

Ah, on vacation 🌴
r? RalfJung or reassign

RalfJung · 2025-04-08T06:44:33Z

compiler/rustc_mir_transform/src/simplify.rs

+//! We must be extremely careful to only apply optimizations that preserve the semantics of the
+//! code, since changes here can affect which programs compile in an insta-stable way.


Removing UB is generally considered "preserving the semantics", since UB programs have "all behaviors". So I think the wording here is potentially confusing.

Suggested change

//! We must be extremely careful to only apply optimizations that preserve the semantics of the

//! code, since changes here can affect which programs compile in an insta-stable way.

//! We must be extremely careful to only apply optimizations that preserve UB and

//! all non-determinism, since changes here can affect which programs compile in

//! an insta-stable way. The normal logic that a program with UB can be changed

//! to do anything does not apply to pre-"runtime" MIR!

RalfJung · 2025-04-08T06:46:06Z

tests/mir-opt/read_from_trivial_switch.rs

This could also use a comment explaining the purpose of the test.

rustbot assigned saethlin Mar 28, 2025

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 Mar 28, 2025

compiler-errors commented Mar 28, 2025

View reviewed changes

This comment has been minimized.

Sign in to view

rustbot 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 Mar 28, 2025

compiler-errors unassigned saethlin Mar 28, 2025

RalfJung reviewed Mar 28, 2025

View reviewed changes

craterbot added S-waiting-on-crater Status: Waiting on a crater run to be completed. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Mar 28, 2025

craterbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-crater Status: Waiting on a crater run to be completed. labels Mar 28, 2025

compiler-errors force-pushed the do-not-optimize-switchint branch from 6bad178 to 01ee75d Compare March 30, 2025 19:01

This comment has been minimized.

Sign in to view

craterbot added S-waiting-on-crater Status: Waiting on a crater run to be completed. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Mar 30, 2025

compiler-errors force-pushed the do-not-optimize-switchint branch from 01ee75d to 6fa280b Compare March 30, 2025 23:21

craterbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-crater Status: Waiting on a crater run to be completed. labels Mar 31, 2025

RalfJung reviewed Apr 1, 2025

View reviewed changes

compiler/rustc_mir_transform/src/simplify.rs Outdated Show resolved Hide resolved

RalfJung reviewed Apr 1, 2025

View reviewed changes

compiler/rustc_mir_transform/src/simplify.rs Show resolved Hide resolved

meithecatte mentioned this pull request Apr 1, 2025

Make closure capturing have consistent and correct behaviour around patterns #138961

Open

compiler-errors force-pushed the do-not-optimize-switchint branch from 6fa280b to 1fed5da Compare April 7, 2025 19:02

Do not optimize out SwitchInt before borrowck, or if Zmir-preserve-ub

e5531b3

compiler-errors force-pushed the do-not-optimize-switchint branch from 1fed5da to e5531b3 Compare April 7, 2025 19:09

rustbot assigned RalfJung Apr 7, 2025

RalfJung reviewed Apr 8, 2025

View reviewed changes

tests/mir-opt/read_from_trivial_switch.rs

Copy link

Member

RalfJung Apr 8, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This could also use a comment explaining the purpose of the test.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Do not remove trivial `SwitchInt` with mir-opt-level=0 #139042

Do not remove trivial `SwitchInt` with mir-opt-level=0 #139042

compiler-errors commented Mar 28, 2025 •

edited

Loading

rustbot commented Mar 28, 2025

rustbot commented Mar 28, 2025

compiler-errors Mar 28, 2025

compiler-errors Mar 28, 2025

tmiasko Mar 28, 2025

compiler-errors Mar 28, 2025

RalfJung Mar 28, 2025

This comment has been minimized.

compiler-errors commented Mar 28, 2025

bors commented Mar 28, 2025

compiler-errors commented Mar 28, 2025

RalfJung Mar 28, 2025

RalfJung Apr 8, 2025

bors commented Mar 28, 2025

oli-obk commented Mar 28, 2025

craterbot commented Mar 28, 2025

tmiasko commented Mar 28, 2025

RalfJung commented Mar 28, 2025

craterbot commented Mar 28, 2025

compiler-errors commented Mar 30, 2025

bors commented Mar 30, 2025

This comment has been minimized.

bors commented Mar 30, 2025

compiler-errors commented Mar 30, 2025

craterbot commented Mar 30, 2025

craterbot commented Mar 30, 2025

craterbot commented Mar 31, 2025

compiler-errors commented Mar 31, 2025

compiler-errors commented Apr 7, 2025

compiler-errors commented Apr 7, 2025

rustbot commented Apr 7, 2025

compiler-errors commented Apr 7, 2025

RalfJung Apr 8, 2025

RalfJung Apr 8, 2025

		//! We must be extremely careful to only apply optimizations that preserve the semantics of the
		//! code, since changes here can affect which programs compile in an insta-stable way.

-//! We must be extremely careful to only apply optimizations that preserve the semantics of the
-//! code, since changes here can affect which programs compile in an insta-stable way.
+//! We must be extremely careful to only apply optimizations that preserve UB and
+//! all non-determinism, since changes here can affect which programs compile in
+//! an insta-stable way. The normal logic that a program with UB can be changed
+//! to do anything does not apply to pre-"runtime" MIR!

Do not remove trivial SwitchInt with mir-opt-level=0 #139042

Are you sure you want to change the base?

Do not remove trivial SwitchInt with mir-opt-level=0 #139042

Conversation

compiler-errors commented Mar 28, 2025 • edited Loading

rustbot commented Mar 28, 2025

rustbot commented Mar 28, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

This comment has been minimized.

compiler-errors commented Mar 28, 2025

bors commented Mar 28, 2025

compiler-errors commented Mar 28, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bors commented Mar 28, 2025

oli-obk commented Mar 28, 2025

craterbot commented Mar 28, 2025

tmiasko commented Mar 28, 2025

RalfJung commented Mar 28, 2025

craterbot commented Mar 28, 2025

compiler-errors commented Mar 30, 2025

bors commented Mar 30, 2025

This comment has been minimized.

bors commented Mar 30, 2025

compiler-errors commented Mar 30, 2025

craterbot commented Mar 30, 2025

craterbot commented Mar 30, 2025

craterbot commented Mar 31, 2025

compiler-errors commented Mar 31, 2025

compiler-errors commented Apr 7, 2025

compiler-errors commented Apr 7, 2025

rustbot commented Apr 7, 2025

compiler-errors commented Apr 7, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Do not remove trivial `SwitchInt` with mir-opt-level=0 #139042

Do not remove trivial `SwitchInt` with mir-opt-level=0 #139042

compiler-errors commented Mar 28, 2025 •

edited

Loading