Deduplicate all files in the wheel cache by charliermarsh · Pull Request #21327 · astral-sh/uv

charliermarsh · 2026-08-27T19:39:53Z

Summary

On main, we support content-addressed caching, but only at the wheel-level. That is, if you download the same wheel twice from different sources, they share a cache entry. But files within or across wheels are not deduplicated at all.

This PR adds deduplication at the file level: every file is now stored under its BLAKE3 hash in a files-v0 bucket. We hardlink these objects into their original locations in archive-v0, so the installation step doesn't change at all -- we're just deduping within the cache (and cache cleanup removes file objects when their hardlink count drops to one).

In the prior proposal (#19694), we included the following table:

File selection	Additional savings	Files hardlinked	Distinct `files-v0` objects
Executables and native libraries	275.7 MiB	3,336	3,088
Any payload file ≥ 10 MiB	235.0 MiB	80	78
Any payload file ≥ 1 MiB	279.7 MiB	373	349
Any payload file ≥ 100 KiB	353.4 MiB	2,830	2,458
Any payload file ≥ 10 KiB	475.7 MiB	23,767	18,722
Any payload file ≥ 1 KiB	537.4 MiB	95,156	66,422
All payload files	545.2 MiB	134,222	87,129

So we're saving 545.2 MiB on my local machine, or about 10% of the cache.

In return, the net effect seems to be something like a <4% slowdown for cold installs (and no effect on warm installs), which I think is probably worthwhile here.

charliemarsh-oai · 2026-08-27T22:02:55Z

N.B. Slop comment for benchmark data.

We benchmarked the optimization in 85c3f7485 against the binary-only cache at a3f977ece, with --preview-features content-addressed-cache enabled on both. These are medians for the complete uv pip install process; positive changes mean slower.

Wheel	Cache	Parent median	Optimized median	Change (95% CI)
AnyIO 4.9.0	cold	90.4 ms	93.4 ms	+3.39% (+2.87% to +3.86%)
AnyIO 4.9.0	warm	29.9 ms	30.3 ms	+1.14% (-1.08% to +3.85%)
SymPy 1.14.0	cold	367.1 ms	381.6 ms	+3.95% (+3.49% to +4.46%)
SymPy 1.14.0	warm	84.6 ms	84.0 ms	-0.64% (-1.04% to -0.11%)
NumPy 2.2.6	cold	314.1 ms	326.5 ms	+3.95% (+3.24% to +4.82%)
NumPy 2.2.6	warm	62.5 ms	62.7 ms	+0.28% (-0.55% to +1.25%)
PyTorch 2.7.1+cpu	cold	2982.6 ms	3072.9 ms	+3.03% (+0.98% to +5.18%)
PyTorch 2.7.1+cpu	warm	404.0 ms	403.1 ms	-0.21% (-0.72% to +0.11%)

Before this optimization, the measured cold regressions were +4.02% for AnyIO 4.9.0, +19.41% for SymPy 1.14.0, +12.75% for NumPy 2.2.6, +15.30% for PyTorch 2.7.1+cpu.

All eight median regressions are below 5%. The PyTorch cold 95% interval still reaches 5.18%, so its upper bound is not below 5%. Every other upper bound is below 5%.

We retained all 1,860 measured installs, including outliers, and combined every confirmation run for this candidate. Cold measurements have 360 paired rounds for AnyIO, 120 each for SymPy and NumPy, and 90 for PyTorch; warm measurements have 60 paired rounds per wheel. Baseline/candidate order alternates, with three warmups per case. Intervals use a paired percentile bootstrap of the ratio of medians with 10,000 resamples.

These measurements ran on Linux/ext4, an AMD EPYC-Milan VM pinned to eight CPUs, with Python 3.12.13. Both binaries use Rust 1.98.0 and the same profiling profile (optimized, no LTO). Each sample installs one pinned local wheel offline, without dependencies or bytecode compilation, using hardlinks into a fresh virtual environment. Cold removes the entire uv cache; warm retains a primed cache. Setup and cleanup are untimed, and the wheel OS page cache is warm. No compilation ran during the benchmarks. These results do not cover other platforms or network-inclusive installs.

All-file deduplication, content/executable identities, the cache layout, complete archives, and copy fallbacks are preserved. Inode checks confirmed that every archived file shares its file-store object for all four wheels. The five targeted integration tests passed ten stress iterations (50 executions), including local and streamed wheels with one and four workers, RECORD handling, cache cleanup, and cross-filesystem installation. Formatting and Clippy with warnings denied also passed.

codspeed-hq · 2026-08-27T23:59:20Z

Merging this PR will not alter performance

✅ 25 untouched benchmarks
⏩ 12 skipped benchmarks¹

_{Comparing charlie/dirhash-all-files (d12d2ad) with main (7c1d80e)}

12 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

astral-sh-bot · 2026-08-28T01:22:56Z

uv test inventory changes

This PR changes the tests when compared with the main base revision.

Added tests: 5
Removed tests: 0
Changed suites: 2

uv-extract: +1 / -0

Added:

uv-extract::dirhash::archive::tests::extracted_file_executable_status

Removed: none

uv::pip_install: +4 / -0

Added:

uv::pip_install::pip_install::all_files_except_record_use_archive_file_store
uv::pip_install::pip_install::binary_payload_copy_fallback_uses_archive_file_store
uv::pip_install::pip_install::binary_payloads_stay_in_archive_without_preview
uv::pip_install::pip_install::binary_payloads_use_archive_file_store

Removed: none

## Summary When content hashing is enabled, we currently allocate and zero a new 64 KiB buffer for every file we copy and hash during streaming extraction. This PR reuses one buffer across the wheel instead. For the PyTorch wheel used in the benchmarks, that reduces buffer allocations for hashing from 11,120 to one, while keeping the buffer size at 64 KiB per active wheel. The following measurements compare #21327 at `a188b8e833aef3c3b4b60a32ed9fafe6ac74186a` with this optimization applied on top, before moving the change onto `main`. They are not measurements against `main`. The Linux benchmarks alternate base and candidate, using pinned wheels served over local HTTP with content-addressed caching enabled: | Cold install | #21327 | #21327 + buffer reuse | Change | | --- | ---: | ---: | ---: | | AnyIO | 110 ms | 107 ms | -2.6% | | SymPy | 845 ms | 775 ms | -8.3% | | NumPy | 627 ms | 567 ms | -9.5% | | PyTorch CPU | 6.50 s | 5.99 s | -7.8% | | 14-package environment, concurrency 4 | 6.95 s | 6.47 s | -7.0% | The individual results above use 16 paired rounds; the full environment uses 12. AnyIO, SymPy, and NumPy were repeated after an initial 20-pair run: the initial AnyIO timings were noisy, while the initial SymPy and NumPy improvements were 7.8% and 6.9%. All original samples were retained. Cached installs and local-wheel controls showed no consistent change. Across the initial runs, repeats, and controls, we measured 672 installs, excluding warmups and cache priming. Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>

charliermarsh · 2026-08-28T17:05:18Z

Okay, I experimented with a bunch of alternatives to try and make this more performant:

Store all files as non-executable, and then apply executable flags at install-time (Store cached wheel permissions in archive manifests #21334, Pipeline file cache publication for streamed wheels #21333). This would allow us to hardlink into files-v0 while we unzip. Unfortunately, it also means we have to copy every executable that we want to install, since hardlinks share a mode. Ultimately, this was a bit tradeoff -- it was slower!
Pre-fetching the central directory so that we can determine whether a given file is executable at extraction-time. (As-is, we only get the central directory after we've unzipped, so we don't know if a file is executable until after it's been extracted.) This also turned out to be slower because we have additional overhead from making more HTTP requests.

I also considered something like "guess whether a file is executable" during extraction (and then, if we're wrong, copy it and change the mode after unzipping). I guess this could end up being more performant, but I wasn't very happy with the heuristics.

Ultimately, I think what we have here is good.

charliermarsh · 2026-08-28T17:08:33Z

If you bundle in #21340, I believe this is also faster than before (i.e., gains from #21340 outweigh the extra cost in this PR).

## Summary Follow-up to #21327. Cache cleanup currently reads the hardlink count of every `files-v0` object separately. On macOS, we can request names, file types, and link counts in batches with `getattrlistbulk`, allocating paths only for files with a single link. Use this fast path for flat cache shards. We keep the existing walk on other platforms, when bulk reads or required attributes aren't available, and for nested directories. We don't follow symlinks, and file removal still uses the existing storage accounting. On macOS, with 87,129 distinct empty file objects across 256 shards and two hardlinks per object, median full-command times across eight paired warm-cache runs were: | Command | Before | After | | ------------------------------ | -----: | -----: | | `uv cache clean <package>` | 386 ms | 105 ms | | `uv cache clean <10 packages>` | 389 ms | 104 ms | | `uv cache prune` | 386 ms | 101 ms | These runs retain every object, so they measure scan cost rather than deletion throughput. The scan is still single-threaded. --------- Co-authored-by: Zanie Blue <contact@zanie.dev>

This PR contains the following updates: | Package | Type | Update | Change | |---|---|---|---| | [ghcr.io/astral-sh/uv](https://github.com/astral-sh/uv) | final | patch | `0.12.7` → `0.12.8` | --- > ⚠️ **Warning** > > Some dependencies could not be looked up. Check the [Dependency Dashboard](issues/51) for more information. --- ### Release Notes <details> <summary>astral-sh/uv (ghcr.io/astral-sh/uv)</summary> ### [`v0.12.8`](https://github.com/astral-sh/uv/blob/HEAD/CHANGELOG.md#0128) Released on 2026-08-31. ##### Enhancements - Warn about invalid tool directories and continue upgrading valid tools with `uv tool upgrade --all` ([#21368](astral-sh/uv#21368)) ##### Preview features - Deduplicate identical files within and across cached wheels with the `content-addressed-cache` preview feature ([#21327](astral-sh/uv#21327)) - Reduce allocations while extracting content-addressed wheels by reusing the hashing buffer across files ([#21340](astral-sh/uv#21340)) - Speed up content-addressed cache cleanup on macOS by reading hard-link counts in bulk ([#21344](astral-sh/uv#21344)) ##### Performance - Prevent concurrent uv processes from downloading and extracting the same remote wheel more than once ([#21379](astral-sh/uv#21379)) - Speed up dependency graph construction from large lockfiles by indexing packages during traversal ([#21373](astral-sh/uv#21373)) - Extend indexed lockfile traversal to exports, dependency trees, audits, and freshness checks ([#21377](astral-sh/uv#21377)) - Speed up warm resolutions by reducing repeated marker interner work ([#21300](astral-sh/uv#21300)) ##### Bug fixes - Do not trust hashes from direct URLs discovered only in wheel metadata when installing with `--require-hashes` ([#21348](astral-sh/uv#21348)) - Use a compatible Azure Storage API version for anonymous and authenticated requests, allowing credential retries when public access is disabled ([#21366](astral-sh/uv#21366)) - Redact Azure shared access signature (`sig`) query parameters from displayed URLs ([#21360](astral-sh/uv#21360)) - Treat projects below one-level workspace member globs as standalone instead of aborting workspace discovery ([#21341](astral-sh/uv#21341)) ##### Other changes - Update `astral-tokio-tar` to 0.7.0 and use effective sizes when tracking extracted hard links ([#21346](astral-sh/uv#21346)) </details> --- ### Configuration 📅 **Schedule**: (UTC) - Branch creation - At any time (no schedule defined) - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Enabled. ♻ **Rebasing**: Whenever PR is behind base branch, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR has been generated by [Mend Renovate CLI](https://github.com/renovatebot/renovate).  Reviewed-on: https://git.tainton.uk/repos/labmcp/pulls/52 Co-authored-by: renovate[bot] <renovate-bot@git.tainton.uk>

Automated mise tool upgrades from local config. mise-managed tools: - `action-validator` - `actionlint` - `aube` - `editorconfig-checker` - `ghalint` - `pinact` - `pipx:gh-action-pulse` - `prek` - `rumdl` - `shellcheck` - `shfmt` - `tombi` - `uv` - `yamlfmt` - `yamllint` - `zizmor` Command: `mise upgrade --bump --local action-validator actionlint aube editorconfig-checker ghalint pinact pipx:gh-action-pulse prek rumdl shellcheck shfmt tombi uv yamlfmt yamllint zizmor` <details> <summary>Version changelog (2 tools)</summary> | Tool | Requested | Installed | |------|-----------|-----------| | `prek` | `0.5.0` → `0.5.1` | `0.5.0` → `0.5.1` | | `uv` | `0.12.7` → `0.12.8` | `0.12.7` → `0.12.8` | </details> <details> <summary>Release notes (2 tools)</summary> <details> <summary>prek: `0.5.0` → `0.5.1` (j178/prek)</summary> ### v0.5.1 ## Release Notes Released on 2026-09-01. ### Enhancements - Add `--hide-status <passed|failed|skipped>` for hook reports ([#2644](j178/prek#2644)) - Add `prek init` for repository setup ([#2636](j178/prek#2636)) - Apply hook `env` during environment creation ([#2650](j178/prek#2650)) - Disable error snippets in `check-yaml` diagnostics ([#2664](j178/prek#2664)) - Show hooks excluded by skip selectors ([#2645](j178/prek#2645)) - Support Pixi for Conda environments ([#2667](j178/prek#2667)) - Support `cargo-binstall` for Rust CLI dependencies ([#2658](j178/prek#2658)) - Warn about unused keys in user settings ([#2665](j178/prek#2665)) ### Bug fixes - Reject unsupported YAML tags in check-yaml ([#2656](j178/prek#2656)) ### Documentation - Clarify local hook documentation ([#2640](j178/prek#2640)) - Clarify pre-commit command compatibility ([#2635](j178/prek#2635)) - Document automatic PR fixes with autofix.ci ([#2643](j178/prek#2643)) - Document check-yaml unsafe support ([#2632](j178/prek#2632)) - Improve setup and workflow documentation ([#2637](j178/prek#2637)) ### Other changes - Drop low-usage release targets ([#2651](j178/prek#2651)) - Generate a prek manifest JSON schema ([#2648](j178/prek#2648)) ### Contributors - @j178 ## Install prek 0.5.1 ### Install prebuilt binaries via shell script ```sh curl --proto '=https' --tlsv1.2 -LsSf https://github.com/j178/prek/releases/download/v0.5.1/prek-installer.sh | sh ``` ### Install prebuilt binaries via powershell script ```sh powershell -ExecutionPolicy Bypass -c "irm https://github.com/j178/prek/releases/download… (truncated) </details> <details> <summary>uv: `0.12.7` → `0.12.8` (astral-sh/uv)</summary> ### 0.12.8 ## Release Notes Released on 2026-08-31. ### Enhancements - Warn about invalid tool directories and continue upgrading valid tools with `uv tool upgrade --all` ([#21368](astral-sh/uv#21368)) ### Preview features - Deduplicate identical files within and across cached wheels with the `content-addressed-cache` preview feature ([#21327](astral-sh/uv#21327)) - Reduce allocations while extracting content-addressed wheels by reusing the hashing buffer across files ([#21340](astral-sh/uv#21340)) - Speed up content-addressed cache cleanup on macOS by reading hard-link counts in bulk ([#21344](astral-sh/uv#21344)) ### Performance - Prevent concurrent uv processes from downloading and extracting the same remote wheel more than once ([#21379](astral-sh/uv#21379)) - Speed up dependency graph construction from large lockfiles by indexing packages during traversal ([#21373](astral-sh/uv#21373)) - Extend indexed lockfile traversal to exports, dependency trees, audits, and freshness checks ([#21377](astral-sh/uv#21377)) - Speed up warm resolutions by reducing repeated marker interner work ([#21300](astral-sh/uv#21300)) ### Bug fixes - Do not trust hashes from direct URLs discovered only in wheel metadata when installing with `--require-hashes` ([#21348](astral-sh/uv#21348)) - Use a compatible Azure Storage API version for anonymous and authenticated requests, allowing credential retries when public access is disabled ([#21366](astral-sh/uv#21366)) - Redact Azure shared access signature (`sig`) query parameters from displayed URLs ([#21360](astral-sh/uv#21360)) - Treat projects below one-level workspace member globs as standalone instead of aborting workspace discovery ([#21341](https://github.com/astra… (truncated) </details> </details> Modified files: - `.mise.toml`

Automated mise tool upgrades from local config. mise-managed tools: - `action-validator` - `actionlint` - `aube` - `editorconfig-checker` - `ghalint` - `pinact` - `pipx:gh-action-pulse` - `prek` - `rumdl` - `shellcheck` - `shfmt` - `tombi` - `uv` - `yamlfmt` - `yamllint` - `zizmor` Command: `mise upgrade --bump --local action-validator actionlint aube editorconfig-checker ghalint pinact pipx:gh-action-pulse prek rumdl shellcheck shfmt tombi uv yamlfmt yamllint zizmor` <details> <summary>Version changelog (2 tools)</summary> | Tool | Requested | Installed | |------|-----------|-----------| | `prek` | `0.5.0` → `0.5.1` | `0.5.0` → `0.5.1` | | `uv` | `0.12.7` → `0.12.8` | `0.12.7` → `0.12.8` | </details> <details> <summary>Release notes (2 tools)</summary> <details> <summary>prek: `0.5.0` → `0.5.1` (j178/prek)</summary> ### v0.5.1 ## Release Notes Released on 2026-09-01. ### Enhancements - Add `--hide-status <passed|failed|skipped>` for hook reports ([#2644](j178/prek#2644)) - Add `prek init` for repository setup ([#2636](j178/prek#2636)) - Apply hook `env` during environment creation ([#2650](j178/prek#2650)) - Disable error snippets in `check-yaml` diagnostics ([#2664](j178/prek#2664)) - Show hooks excluded by skip selectors ([#2645](j178/prek#2645)) - Support Pixi for Conda environments ([#2667](j178/prek#2667)) - Support `cargo-binstall` for Rust CLI dependencies ([#2658](j178/prek#2658)) - Warn about unused keys in user settings ([#2665](j178/prek#2665)) ### Bug fixes - Reject unsupported YAML tags in check-yaml ([#2656](j178/prek#2656)) ### Documentation - Clarify local hook documentation ([#2640](j178/prek#2640)) - Clarify pre-commit command compatibility ([#2635](j178/prek#2635)) - Document automatic PR fixes with autofix.ci ([#2643](j178/prek#2643)) - Document check-yaml unsafe support ([#2632](j178/prek#2632)) - Improve setup and workflow documentation ([#2637](j178/prek#2637)) ### Other changes - Drop low-usage release targets ([#2651](j178/prek#2651)) - Generate a prek manifest JSON schema ([#2648](j178/prek#2648)) ### Contributors - @j178 ## Install prek 0.5.1 ### Install prebuilt binaries via shell script ```sh curl --proto '=https' --tlsv1.2 -LsSf https://github.com/j178/prek/releases/download/v0.5.1/prek-installer.sh | sh ``` ### Install prebuilt binaries via powershell script ```sh powershell -ExecutionPolicy Bypass -c "irm https://github.com/j178/prek/releases/download… (truncated) </details> <details> <summary>uv: `0.12.7` → `0.12.8` (astral-sh/uv)</summary> ### 0.12.8 ## Release Notes Released on 2026-08-31. ### Enhancements - Warn about invalid tool directories and continue upgrading valid tools with `uv tool upgrade --all` ([#21368](astral-sh/uv#21368)) ### Preview features - Deduplicate identical files within and across cached wheels with the `content-addressed-cache` preview feature ([#21327](astral-sh/uv#21327)) - Reduce allocations while extracting content-addressed wheels by reusing the hashing buffer across files ([#21340](astral-sh/uv#21340)) - Speed up content-addressed cache cleanup on macOS by reading hard-link counts in bulk ([#21344](astral-sh/uv#21344)) ### Performance - Prevent concurrent uv processes from downloading and extracting the same remote wheel more than once ([#21379](astral-sh/uv#21379)) - Speed up dependency graph construction from large lockfiles by indexing packages during traversal ([#21373](astral-sh/uv#21373)) - Extend indexed lockfile traversal to exports, dependency trees, audits, and freshness checks ([#21377](astral-sh/uv#21377)) - Speed up warm resolutions by reducing repeated marker interner work ([#21300](astral-sh/uv#21300)) ### Bug fixes - Do not trust hashes from direct URLs discovered only in wheel metadata when installing with `--require-hashes` ([#21348](astral-sh/uv#21348)) - Use a compatible Azure Storage API version for anonymous and authenticated requests, allowing credential retries when public access is disabled ([#21366](astral-sh/uv#21366)) - Redact Azure shared access signature (`sig`) query parameters from displayed URLs ([#21360](astral-sh/uv#21360)) - Treat projects below one-level workspace member globs as standalone instead of aborting workspace discovery ([#21341](https://github.com/astra… (truncated) </details> </details> Modified files: - `.mise.toml` Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

This PR contains the following updates: | Package | Type | Update | Change | |---|---|---|---| | [uv](https://github.com/astral-sh/uv) | tools | patch | `0.12.7` → `0.12.9` | --- > ⚠️ **Warning** > > Some dependencies could not be looked up. Check the [Dependency Dashboard](issues/141) for more information. --- ### Release Notes <details> <summary>astral-sh/uv (uv)</summary> ### [`v0.12.9`](https://github.com/astral-sh/uv/blob/HEAD/CHANGELOG.md#0129) [Compare Source](astral-sh/uv@0.12.8...0.12.9) Released on 2026-09-01. ##### Python - Add CPython 3.15.0rc2 ([#21413](astral-sh/uv#21413), [#21415](astral-sh/uv#21415)) ##### Enhancements - Add `--no-locked` and `--no-frozen` to disable lock modes enabled by `UV_LOCKED` and `UV_FROZEN` for a single invocation ([#21408](astral-sh/uv#21408)) - Report the exact command-line lock-mode flag in warnings and errors ([#21402](astral-sh/uv#21402)) ##### Performance - Speed up cold wheel installs by extracting each streaming ZIP archive in a single blocking task and reusing buffers across files ([#21372](astral-sh/uv#21372)) ##### Bug fixes - Update `async_http_range_reader` to 0.11.1 to address a potential memory-safety issue when reading metadata ranges from untrusted wheels ([#21401](astral-sh/uv#21401)) - Remove sensitive headers when redirects cross authentication realms, including same-host redirects that change URL schemes ([#21382](astral-sh/uv#21382)) - Redact secrets in signed URLs from retry diagnostics, including nested request errors ([#21381](astral-sh/uv#21381)) - Give `--locked`, `--frozen`, `--check`, and `--check-exists` precedence over conflicting `UV_LOCKED` and `UV_FROZEN` values ([#21396](astral-sh/uv#21396)) - Prevent concurrent uv processes from redundantly extracting the same local or source-built wheel ([#21400](astral-sh/uv#21400)) ### [`v0.12.8`](https://github.com/astral-sh/uv/blob/HEAD/CHANGELOG.md#0128) [Compare Source](astral-sh/uv@0.12.7...0.12.8) Released on 2026-08-31. ##### Enhancements - Warn about invalid tool directories and continue upgrading valid tools with `uv tool upgrade --all` ([#21368](astral-sh/uv#21368)) ##### Preview features - Deduplicate identical files within and across cached wheels with the `content-addressed-cache` preview feature ([#21327](astral-sh/uv#21327)) - Reduce allocations while extracting content-addressed wheels by reusing the hashing buffer across files ([#21340](astral-sh/uv#21340)) - Speed up content-addressed cache cleanup on macOS by reading hard-link counts in bulk ([#21344](astral-sh/uv#21344)) ##### Performance - Prevent concurrent uv processes from downloading and extracting the same remote wheel more than once ([#21379](astral-sh/uv#21379)) - Speed up dependency graph construction from large lockfiles by indexing packages during traversal ([#21373](astral-sh/uv#21373)) - Extend indexed lockfile traversal to exports, dependency trees, audits, and freshness checks ([#21377](astral-sh/uv#21377)) - Speed up warm resolutions by reducing repeated marker interner work ([#21300](astral-sh/uv#21300)) ##### Bug fixes - Do not trust hashes from direct URLs discovered only in wheel metadata when installing with `--require-hashes` ([#21348](astral-sh/uv#21348)) - Use a compatible Azure Storage API version for anonymous and authenticated requests, allowing credential retries when public access is disabled ([#21366](astral-sh/uv#21366)) - Redact Azure shared access signature (`sig`) query parameters from displayed URLs ([#21360](astral-sh/uv#21360)) - Treat projects below one-level workspace member globs as standalone instead of aborting workspace discovery ([#21341](astral-sh/uv#21341)) ##### Other changes - Update `astral-tokio-tar` to 0.7.0 and use effective sizes when tracking extracted hard links ([#21346](astral-sh/uv#21346)) </details> --- ### Configuration 📅 **Schedule**: (in timezone Europe/London) - Branch creation - At any time (no schedule defined) - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR has been generated by [Mend Renovate CLI](https://github.com/renovatebot/renovate).  Reviewed-on: https://forgejo.hayden.moe/hayden/phoebe/pulls/430

charliermarsh temporarily deployed to automations August 27, 2026 19:40 — with GitHub Actions Inactive

astral-automations-bot Bot mentioned this pull request Aug 27, 2026

Linux cargo-test setup fails when Depot runners lack Minix support astral-sh/uv-dev#896

Open

charliermarsh temporarily deployed to automations August 27, 2026 22:01 — with GitHub Actions Inactive

charliermarsh temporarily deployed to automations August 27, 2026 23:35 — with GitHub Actions Inactive

charliermarsh marked this pull request as ready for review August 28, 2026 00:03

charliermarsh force-pushed the charlie/dirhash-binaries branch from a3f977e to 1f28fd5 Compare August 28, 2026 00:08

charliermarsh force-pushed the charlie/dirhash-all-files branch from 0a8edc0 to a108fab Compare August 28, 2026 00:08

charliermarsh temporarily deployed to automations August 28, 2026 00:09 — with GitHub Actions Inactive

charliermarsh changed the base branch from charlie/dirhash-binaries to main August 28, 2026 00:09

charliermarsh added enhancement New feature or improvement to existing functionality preview Experimental behavior labels Aug 28, 2026

charliermarsh requested review from woodruffw and zanieb August 28, 2026 00:13

charliermarsh temporarily deployed to automations August 28, 2026 00:23 — with GitHub Actions Inactive

charliermarsh mentioned this pull request Aug 28, 2026

Pipeline file cache publication for streamed wheels #21333

Draft

charliermarsh mentioned this pull request Aug 28, 2026

Store cached wheel permissions in archive manifests #21334

Draft

zsol self-requested a review August 28, 2026 09:53

This was referenced Aug 28, 2026

Prefetch wheel central directories for streamed file publication #21338

Draft

Pipeline wheel file publication with executable deferral #21339

Draft

charliemarsh-oai mentioned this pull request Aug 28, 2026

Reuse the hashing buffer across streamed wheel files #21340

Merged

charliermarsh force-pushed the charlie/dirhash-all-files branch from a188b8e to 694af80 Compare August 28, 2026 16:56

charliermarsh had a problem deploying to automations August 28, 2026 16:57 — with GitHub Actions Error

charliermarsh temporarily deployed to automations August 28, 2026 17:11 — with GitHub Actions Inactive

charliermarsh temporarily deployed to automations August 28, 2026 17:28 — with GitHub Actions Inactive

charliermarsh and others added 12 commits August 31, 2026 08:56

Separate file-cache deduplication from installation

65f6c62

Keep cache-only callers and generated schema consistent

d0682a4

Model hashed and unhashed extraction explicitly

b6cdda1

Bundle hashed wheel metadata and trim helper tests

8acb8c8

Remove archive publication comment

7c10407

Avoid duplicate extracted wheel file records

390db4e

Deduplicate all files in the wheel cache

7818a02

Optimize wheel file cache publication

ef2a9b7

Exclude RECORD from wheel file deduplication

d2e340c

Include executable status in file cache hashes

b37fdee

Prune orphaned file objects when cleaning an unreferenced cache

af30c8c

Prune file objects once per cache clean

87bd56f

charliermarsh force-pushed the charlie/dirhash-all-files branch from 73c621b to 87bd56f Compare August 31, 2026 12:59

charliermarsh temporarily deployed to automations August 31, 2026 13:00 — with GitHub Actions Inactive

Address wheel cache review feedback

d12d2ad

charliermarsh temporarily deployed to automations August 31, 2026 13:17 — with GitHub Actions Inactive

charliermarsh mentioned this pull request Aug 31, 2026

Reuse cached small files during wheel extraction #21375

Closed

charliermarsh merged commit 15bf04f into main Aug 31, 2026
82 checks passed

charliermarsh deleted the charlie/dirhash-all-files branch August 31, 2026 13:28

Foxhunt mentioned this pull request Aug 31, 2026

Daily Hacker News 31-08-2026 Foxhunt/daily-hackernews#783

Open

github-actions Bot mentioned this pull request Aug 31, 2026

[Trend Digest] Week of 2026-08-31 strikersam/autonomous-ai-agency#1397

Closed

blka mentioned this pull request Aug 31, 2026

Daily Hacker News 31/08/2026 blka/daily-hackernews#530

Open

BrewTestBot mentioned this pull request Sep 1, 2026

uv 0.12.8 Homebrew/homebrew-core#301648

Merged

github-actions Bot mentioned this pull request Sep 1, 2026

[2026-08-31] “I just chose words carefully” jiacai2050/mofish#1565

Open

ranger-ross mentioned this pull request Sep 1, 2026

Per-user compiled artifact cache rust-lang/cargo#5931

Open

9 tasks

Navigation Menu

Uh oh!

Deduplicate all files in the wheel cache - #21327

Deduplicate all files in the wheel cache#21327
charliermarsh merged 24 commits into
mainfrom
charlie/dirhash-all-files

charliermarsh commented Aug 27, 2026 •

edited

Loading

Uh oh!

charliemarsh-oai commented Aug 27, 2026 •

edited by charliermarsh

Loading

Uh oh!

codspeed-hq Bot commented Aug 27, 2026 •

edited

Loading

Uh oh!

astral-sh-bot Bot commented Aug 28, 2026

Uh oh!

charliermarsh commented Aug 28, 2026

Uh oh!

charliermarsh commented Aug 28, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Uh oh!

Conversation

charliermarsh commented Aug 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Uh oh!

charliemarsh-oai commented Aug 27, 2026 • edited by charliermarsh Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codspeed-hq Bot commented Aug 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging this PR will not alter performance

Footnotes

Uh oh!

astral-sh-bot Bot commented Aug 28, 2026

uv test inventory changes

Uh oh!

charliermarsh commented Aug 28, 2026

Uh oh!

charliermarsh commented Aug 28, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

charliermarsh commented Aug 27, 2026 •

edited

Loading

charliemarsh-oai commented Aug 27, 2026 •

edited by charliermarsh

Loading

codspeed-hq Bot commented Aug 27, 2026 •

edited

Loading