Dogfood `-Zno-embed-metadata` in the standard library #145343

Kobzol · 2025-08-13T09:11:08Z

This PR dogfoods the -Zno-embed-metadata flag in the standard library. This removes the .rmeta portion out of the libstd.so file, thus reducing its filesize on disk. Notably, this reduces the amount of MiB that we ship to people who download the standard library.

I think that the only way to find out what this breaks is to try to run full CI, and then try to land it on nightly :)

r? @ghost

Kobzol · 2025-08-13T09:50:44Z

@bors try

Dogfood `-Zno-embed-metadata` in the compiler

rust-bors · 2025-08-13T12:04:01Z

☀️ Try build successful (CI)
Build commit: 6edf4f5 (6edf4f50158d893f8aa51ea5a04ba02370776b95, parent: b1b26b834d85e84b46aa8f8f3ce210a1627aa85f)

Kobzol · 2025-08-13T12:16:26Z

@rust-timer build 6edf4f5

rust-timer · 2025-08-13T16:27:06Z

Finished benchmarking commit (6edf4f5): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +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)	0.7%	[0.2%, 1.2%]	7
Regressions ❌ (secondary)	0.4%	[0.1%, 0.9%]	43
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-0.2%	[-0.5%, -0.0%]	3
All ❌✅ (primary)	0.7%	[0.2%, 1.2%]	7

Max RSS (memory usage)

Results (primary -0.7%, secondary -0.8%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	1.3%	[1.3%, 1.3%]	1
Regressions ❌ (secondary)	2.7%	[2.3%, 3.2%]	4
Improvements ✅ (primary)	-1.4%	[-2.3%, -0.5%]	3
Improvements ✅ (secondary)	-2.5%	[-5.2%, -0.8%]	8
All ❌✅ (primary)	-0.7%	[-2.3%, 1.3%]	4

Cycles

Results (primary 3.6%, secondary 3.7%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	3.6%	[3.6%, 3.6%]	1
Regressions ❌ (secondary)	3.7%	[2.0%, 5.3%]	4
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	3.6%	[3.6%, 3.6%]	1

Binary size

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

Bootstrap: 464.328s -> 465.28s (0.21%)
Artifact size: 377.44 MiB -> 377.37 MiB (-0.02%)

bors · 2025-08-14T00:07:36Z

☔ The latest upstream changes (presumably #145366) made this pull request unmergeable. Please resolve the merge conflicts.

Kobzol · 2025-08-25T13:30:58Z

@bors try @rust-timer queue

Dogfood `-Zno-embed-metadata` in the compiler

Kobzol · 2025-08-25T15:04:19Z

@bors try @rust-timer queue

Dogfood `-Zno-embed-metadata` in the compiler

Kobzol · 2025-08-25T17:08:19Z

@bors try @rust-timer queue

rust-bors · 2026-01-12T12:08:58Z

☀️ Try build successful (CI)
Build commit: 548971f (548971f554342047d492efc01f0161c8309d3845, parent: 561364e4d5ccc506f610208a4989e91fdbdc8ca7)

bjorn3 · 2026-01-12T12:10:50Z

@bors r+

rust-bors · 2026-01-12T12:10:55Z

📌 Commit e03cb1f has been approved by bjorn3

It is now in the queue for this repository.

@ghost

Dogfood `-Zno-embed-metadata` in the standard library This PR dogfoods the [`-Zno-embed-metadata`](rust-lang/cargo#15495) flag in the standard library. This removes the .rmeta portion out of the `libstd.so` file, thus reducing its filesize on disk. Notably, this reduces the amount of MiB that we ship to people who download the standard library. I think that the only way to find out what this breaks is to try to run full CI, and then try to land it on nightly :) r? @ghost

@ghost

…uwer Rollup of 13 pull requests Successful merges: - #145343 (Dogfood `-Zno-embed-metadata` in the standard library) - #150151 (Destabilise `target-spec-json`) - #150723 (std: move `errno` and related functions into `sys::io`) - #150771 (Remove legacy homu `try` and `auto` branch mentions) - #150826 (Add `f16` inline ASM support for s390x) - #150934 (Move some checks from `check_doc_attrs` directly into `rustc_attr_parsing`) - #150943 (Port `#[must_not_suspend]` to attribute parser) - #150990 (std: sys: net: uefi: Make TcpStream Send) - #150995 (core: ptr: split_at_mut: fix typo in safety doc) - #150998 (Relax test expectation for @__llvm_profile_runtime_user) - #151002 (Remove a workaround for a bug (take 2)) - #151005 (Fix typo in `MaybeUninit` docs) - #151011 (Update books) r? @ghost

@ghost

Dogfood `-Zno-embed-metadata` in the standard library This PR dogfoods the [`-Zno-embed-metadata`](rust-lang/cargo#15495) flag in the standard library. This removes the .rmeta portion out of the `libstd.so` file, thus reducing its filesize on disk. Notably, this reduces the amount of MiB that we ship to people who download the standard library. I think that the only way to find out what this breaks is to try to run full CI, and then try to land it on nightly :) r? @ghost

@ghost

Rollup of 12 pull requests Successful merges: - #145343 (Dogfood `-Zno-embed-metadata` in the standard library) - #150151 (Destabilise `target-spec-json`) - #150723 (std: move `errno` and related functions into `sys::io`) - #150826 (Add `f16` inline ASM support for s390x) - #150934 (Move some checks from `check_doc_attrs` directly into `rustc_attr_parsing`) - #150943 (Port `#[must_not_suspend]` to attribute parser) - #150990 (std: sys: net: uefi: Make TcpStream Send) - #150995 (core: ptr: split_at_mut: fix typo in safety doc) - #150998 (Relax test expectation for @__llvm_profile_runtime_user) - #151002 (Remove a workaround for a bug (take 2)) - #151005 (Fix typo in `MaybeUninit` docs) - #151011 (Update books) r? @ghost

Kobzol · 2026-01-13T07:07:10Z

In case the current rollup doesn't make it, I'd rather mark this as rollup=never, hard to say if this change will pass all tests.

@bors rollup=never

rust-bors · 2026-01-13T10:46:44Z

☀️ Test successful - CI
Approved by: bjorn3
Pushing b2a322b to main...

github-actions · 2026-01-13T10:49:21Z

What is this?

This is an experimental post-merge analysis report that shows differences in test outcomes between the merged PR and its parent PR.

Comparing 2f1bd3f (parent) -> b2a322b (this PR)

Test differences

Show 5 test diffs

Stage 2

[debuginfo-cdb] tests/debuginfo/numeric-types.rs: ignore (ignored when the CDB version is lower than 10.0.26100.2161) -> pass (J0)
[debuginfo-cdb] tests/debuginfo/range-types.rs: ignore (ignored when the CDB version is lower than 10.0.26100.2161) -> pass (J0)
[debuginfo-cdb] tests/debuginfo/unit-type.rs: ignore (ignored when the CDB version is lower than 10.0.26100.2161) -> pass (J0)

Additionally, 2 doctest diffs were found. These are ignored, as they are noisy.

Job group index

J0: aarch64-msvc-1

Test dashboard

Run

cargo run --manifest-path src/ci/citool/Cargo.toml -- \
    test-dashboard b2a322beb29110e22a1782e2ce5ed2a0719b81ed --output-dir test-dashboard

And then open test-dashboard/index.html in your browser to see an overview of all executed tests.

Job duration changes

aarch64-apple: 12590.2s -> 9255.9s (-26.5%)
x86_64-gnu-gcc: 3574.7s -> 3032.7s (-15.2%)
x86_64-rust-for-linux: 3164.5s -> 2688.2s (-15.1%)
pr-check-1: 1960.2s -> 1685.4s (-14.0%)
dist-aarch64-apple: 8773.0s -> 7623.4s (-13.1%)
i686-gnu-2: 6006.5s -> 5259.7s (-12.4%)
pr-check-2: 2651.3s -> 2328.5s (-12.2%)
i686-gnu-1: 8339.6s -> 7365.6s (-11.7%)
aarch64-gnu-llvm-20-2: 3042.7s -> 2705.8s (-11.1%)
aarch64-msvc-2: 5990.0s -> 6653.3s (+11.1%)

How to interpret the job duration changes?

Job durations can vary a lot, based on the actual runner instance
that executed the job, system noise, invalidated caches, etc. The table above is provided
mostly for t-infra members, for simpler debugging of potential CI slow-downs.

rust-timer · 2026-01-13T11:28:54Z

Finished benchmarking commit (b2a322b): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Our benchmarks found a performance regression caused by this PR.
This might be an actual regression, but it can also be just noise.

Next Steps:

If the regression was expected or you think it can be justified,
please write a comment with sufficient written justification, and add
@rustbot label: +perf-regression-triaged to it, to mark the regression as triaged.
If you think that you know of a way to resolve the regression, try to create
a new PR with a fix for the regression.
If you do not understand the regression or you think that it is just noise,
you can ask the @rust-lang/wg-compiler-performance working group for help (members of this group
were already notified of this PR).

@rustbot label: +perf-regression
cc @rust-lang/wg-compiler-performance

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)	0.9%	[0.5%, 1.1%]	5
Regressions ❌ (secondary)	0.3%	[0.1%, 0.8%]	45
Improvements ✅ (primary)	-0.3%	[-0.3%, -0.2%]	2
Improvements ✅ (secondary)	-0.0%	[-0.0%, -0.0%]	1
All ❌✅ (primary)	0.5%	[-0.3%, 1.1%]	7

Max RSS (memory usage)

Results (secondary -0.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.5%	[1.7%, 3.1%]	3
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.1%	[-3.1%, -1.2%]	4
All ❌✅ (primary)	-	-	0

Cycles

Results (secondary -3.4%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.4%	[-4.1%, -2.7%]	2
All ❌✅ (primary)	-	-	0

Binary size

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

Bootstrap: 475.39s -> 473.958s (-0.30%)
Artifact size: 391.43 MiB -> 383.19 MiB (-2.11%)

Kobzol · 2026-01-13T11:30:26Z

The tiny regressions are only noticeable on the smallest crates, they are caused by now loading both the .rmeta and .rlib files of stdlib.

@rustbot label: +perf-regression-triaged

bjorn3 · 2026-01-13T19:43:14Z

This saved enough disk space that the used column of df -h went from 239GB to 238GB. Probably didn't save a full GB though.

Kobzol · 2026-01-13T19:45:57Z

I posted some statistics for the libstd component here (GitHub hid it in the meantime...).

bjorn3 · 2026-01-13T19:50:04Z

The savings to the amount of space it takes to build the standard library should be much larger than the savings of the final artifact size as the latter already skipped the duplicate crate metadata for everything other than libstd.so.

Kobzol · 2026-01-13T19:53:46Z

Right, but we don't really care that much about the total disk savings on CI, I guess? Locally it might be more important, but probably still ~negligent. The size of the rustup-distributed toolchain is the win I'm after here.

That is true for rustc components, of course. For normal Rust code, -Zno-embed-metadata will ofc have nice wins for disk space. But here I'm doing it mostly for network bandwidth savings.

bjorn3 · 2026-01-13T20:10:58Z

For CI it could potentially help a bit with being able to build on smaller runners with less free space. Locally for most people it wouldn't matter much, but on dev-desktop it would make me need to throw away the build directory less frequently and I'm sure there are people who need/want to build rustc in relatively constrained environments too. And I believe some people have multiple worktrees checked out that each take up a lot of space. This PR likely won't help all that much with that, but my PR for doing the same for rustc should help a non-trivial amount.

rust-bors bot added a commit that referenced this pull request Aug 13, 2025

Auto merge of #145343 - Kobzol:dogfood-z-no-embed-metadata, r=<try>

6edf4f5

Dogfood `-Zno-embed-metadata` in the compiler