Revert "miri: make sure we can find link_section statics even for the local crate"

[lqd]

This PR reverts #126938 as requested by its author, to fix the #127052 regression.

Fixes #127052

We should probably improve the used rmake test(s) in the future, but this should do for now.

[lqd]

@bors try

[bors]

⌛ Trying commit 57931e5 with merge 6ee8465...

[workingjubilee]

Thank you for filing this revert + regression test!

[compiler-errors]

thanks @lqd, r=me when this is done bors-trying so we can fix this

[lqd]

The test technically looks good on the try build already.

test [ui] tests\ui\linkage-attr\unreferenced-used-static-issue-127052.rs ... ok

(and by now an old windows box I have has finished rebuilding LLVM -- to workaround the recent linking issues that other people currently also experience when building rustc -- so I also checked that the test does fail on master without the revert, no need for another try build 🚀).

I'm not sure if //@ only-msvc is restrictive enough in this case (but probably?), and I guess the test comment could be more explicit about what happens but nbd -- so I wanted to ask @ChrisDenton but it seems fine for a revert :)

And if this is indeed what caused CI to fail recently maybe we can bump the prio.

[rustbot]

The Miri subtree was changed

cc @rust-lang/miri

[matthiaskrgr]

@bors p=25 because this problem broke a bunch of ci runs already

[bors]

☀️ Try build successful - checks-actions
Build commit: 6ee8465 (6ee8465a2ca6b4af09e56482223dcfacd262ffa4)

[lqd]

finallyyyyyyy 😴

@bors r=compiler-errors

[bors]

📌 Commit 57931e5 has been approved by compiler-errors

It is now in the queue for this repository.

[bors]

⌛ Testing commit 57931e5 with merge 9ed2ab3...

[rust-highfive]

📣 Toolstate changed by #127099!

Tested on commit 9ed2ab3.
Direct link to PR: #127099

🎉 reference on windows: test-fail → test-pass (cc @Havvy @ehuss @matthewjasper).

[bors]

☀️ Test successful - checks-actions
Approved by: compiler-errors
Pushing 9ed2ab3 to master...

[rust-timer]

Finished benchmarking commit (9ed2ab3): comparison URL.

Overall result: ✅ improvements - no action needed

@rustbot label: -perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-1.8%	[-1.8%, -1.8%]	1
All ❌✅ (primary)	-	-	0

Max RSS (memory usage)

Results (primary -1.7%, secondary 4.8%)

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	1.7%	[0.7%, 2.7%]	2
Regressions ❌ (secondary)	4.8%	[4.8%, 4.8%]	1
Improvements ✅ (primary)	-8.6%	[-8.6%, -8.6%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-1.7%	[-8.6%, 2.7%]	3

Cycles

This benchmark run did not return any relevant results for this metric.

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 694.621s -> 695.135s (0.07%)
Artifact size: 324.60 MiB -> 324.56 MiB (-0.01%)

[RalfJung]

Why shouldn't we run this test on all targets?

[lqd]

The specific issue doesn't exist elsewhere so it's less useful there: the existing tests already cover #[used] features and targets more than this test. Of course, it wouldn't hurt. Feel free to change the directive to something you prefer.

[RalfJung]

Well, clearly the existing tests do not cover #[used] in the binary, or my PR would not have landed.

[apiraino]

thank you @lqd 🙇‍♂️

(sometimes I have to spell it out where a emoji is not enough)