native_lib/trace: fix and reenable #4435

nia-e · 2025-07-02T00:56:42Z

Per discussion in various PRs (#4401, #4405), this addresses (some of) the points raised. Also using this to check if CI passes on all architectures.

nia-e · 2025-07-02T03:31:37Z

The only options I see for making the builds work nicely regardless of arch are either (a) copy-pasting a massive cfg macro everywhere; (b) copy-pasting said massive macro only a couple times to declare some stub functions; or (c) give in and add a small build script. Given the size of Miri I assume the choice to not have said build script was very deliberate, so I can try to do one of the other options (or maybe I'm missing something!) but for now this hopefully will build. This would have solved everything trivially, but oh well

RalfJung · 2025-07-02T07:09:32Z

It should be entirely possible to do this with a single cfg that decides whether we include the "real" implementation of native_lib::tracing, or a fallback.

Cargo.toml

src/bin/miri.rs

RalfJung · 2025-07-02T13:18:02Z

src/shims/native_lib/mod.rs

-//#[cfg(target_os = "linux")]
-//pub mod trace;
+#[cfg(trace)]
+pub mod trace;


To reduce the amount of cfg in this file, I would suggest something like this:

#[cfg_attr(not(all(target_os = "linux", any(target_arch = "x86_64", ...), path = "trace/stub.rs")] mod trace;

and then the stub.rs file contains just a bunch of empty functions and dummy type aliases so that to the outside, it looks like all the same functions and types exist.

RalfJung · 2025-07-02T13:18:56Z

src/alloc/mod.rs

@@ -1,5 +1,5 @@
 mod alloc_bytes;
-#[cfg(target_os = "linux")]


Let's just keep using that allocator on all Linux targets, no harm in that.

src/bin/miri.rs

src/alloc/isolated_alloc.rs

RalfJung · 2025-07-05T07:58:16Z

@rustbot author

rustbot · 2025-07-05T07:58:19Z

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

src/shims/native_lib/mod.rs

src/shims/native_lib/trace/mod.rs

src/shims/native_lib/mod.rs

src/shims/native_lib/trace/child.rs

src/shims/native_lib/trace/mod.rs

src/bin/miri.rs

src/shims/native_lib/trace/stub.rs

RalfJung · 2025-07-07T12:06:12Z

@rustbot ready
(I know this is redundant but I need something in my inbox to represent this ;)

Co-authored-by: Ralf Jung <post@ralfj.de>

RalfJung

I have rebased and pushed a commit with some cleanups, please check if those make sense.

RalfJung · 2025-07-08T20:17:14Z

src/shims/native_lib/trace/parent.rs

@@ -417,71 +412,13 @@ fn handle_segfault(
                            if acc_ty.is_writable() {


Can it actually happen that the same operation is both a read and a write? Is that used for instance for atomic fetch-and-add and operations like that?

nia-e force-pushed the trace-fixups branch 4 times, most recently from 37fbac4 to d2ed49d Compare July 2, 2025 03:27

nia-e force-pushed the trace-fixups branch 2 times, most recently from 0110f5a to 6907e9d Compare July 2, 2025 03:58

RalfJung reviewed Jul 2, 2025

View reviewed changes

src/alloc/isolated_alloc.rs Outdated Show resolved Hide resolved

rustbot added the S-waiting-on-author Status: Waiting for the PR author to address review comments label Jul 5, 2025

nia-e force-pushed the trace-fixups branch 2 times, most recently from 3b1584e to 93015d6 Compare July 5, 2025 10:09