René's URL Explorer Experiment


Title: gh-161478: Collect large compression outputs in blocks again, except with glibc by maikschulze · Pull Request #161482 · python/cpython · GitHub

Open Graph Title: gh-161478: Collect large compression outputs in blocks again, except with glibc by maikschulze · Pull Request #161482 · python/cpython

X Title: gh-161478: Collect large compression outputs in blocks again, except with glibc by maikschulze · Pull Request #161482 · python/cpython

Description: Fixes gh-161478: decompression into large outputs got slower in 3.15 on macOS and Windows, from a few MiB of output (4 on macOS, 16 on Windows), after f262297 (gh-139877) moved _BlocksOutputBuffer to PyBytesWriter. Since f262297, _BlocksOutputBuffer grows one PyBytesWriter with realloc(). What that costs depends on the allocator: Linux (glibc): realloc() grows a large block in place with mremap(). zstd decompression is faster than in 3.14 and the other modules are within about 15%, as f262297 measured. macOS: realloc() grows a large block mostly in place too, but into address space the process has never touched, so every call faults in every 16 KiB page of its output. 3.14's fixed-size blocks were reused by malloc from call to call, with no page faults. The freed block isn't reused either, and the allocator keeps it, so the process footprint grows with every call (below). The same happens with PYTHONMALLOC=mimalloc (zstd 2.2–3.3× slower than 3.14, gzip 1.3–1.5×), so this isn't a quirk of Apple's malloc: what costs is that a block grown with realloc() is never reused by the next call, while fixed-size blocks are. Windows: the process heap behind the UCRT's malloc returns large blocks to the OS on free, and the growing writer, over-allocated by 50%, faults in more fresh memory than 3.14's blocks did: 3.3 page faults per output page against 2.0–2.1, and each realloc() copies the whole capacity. The fix stays in this header rather than in PyBytesWriter: the writer's contract is one contiguous buffer, and a join is only cheap when the consumer can write into separate blocks, as these modules can. Change The code change is in Include/internal/pycore_blocks_output_buffer.h only; the modules and the buffer's API stay as they are. The other files are the NEWS entry and tests. The head, the first 1.34 MiB (the first four blocks of the size sequence, 32 KiB + 64 KiB + 256 KiB + 1 MiB), still grows the PyBytesWriter. An output that fits is returned without a join, so small outputs keep 3.15's gain over 3.14. A buffer that starts at another size, as zlib.decompress() does with bufsize, grows in the writer until the next block would take it past 1.34 MiB; one that starts above that, such as zstd's compressBound() start, switches to blocks at its first growth. The writer keeps its own over-allocation (25%, 50% on Windows), so up to 0.34 MiB (0.67 MiB) of its capacity goes unused after the switch. Later blocks are separate bytes objects in a list, joined once by _BlocksOutputBuffer_Finish(), as in 3.14. An output that ends exactly at the head is still returned from the writer. This is the behaviour everywhere the writer isn't known to grow in place: macOS and Windows, where it was measured; platforms that weren't (FreeBSD, …); and allocators that copy on realloc(), Android's and mimalloc in the free-threaded build, which with PYTHONMALLOC=mimalloc measured like macOS. They all get back what 3.14 did. With glibc, in the default build, OUTPUT_BUFFER_HEAD_SIZE is PY_SSIZE_T_MAX, so the writer grows for all blocks, as now; the list path is unreachable there. Without that exception, zstd decompression on Linux got 1.3–2.3× slower than main from 4 MiB (release builds without PGO): the join costs more than mremap(). The condition is compile-time, so it can't see PYTHONMALLOC=mimalloc on such a build; the writer keeps growing there. Results main at 7054b71 against this branch, both PGO (with LTO on macOS), with python-build-standalone 3.14.10 for reference. bench_compression.py from gh-161478, run alternately three times per interpreter, best of three; one-shot decompression of the reproducer's data. macOS 26.7 on an Apple M4 Pro; Windows 11 on a Ryzen 9 7950X (MSVC 14.41, PGO without LTO). The change in time with this PR (− is faster), against main and against 3.14.10; the times themselves are in the collapsed tables below. Bold marks 15% or more either way. The noise level is about ±10%: on Linux, where the two binaries run the same code for this path, they differ by up to 12%, with the signs flipping between a PGO pair and a non-PGO pair of builds. time with this PR relative to main 1 MiB 4 MiB 16 MiB 64 MiB macOS arm64, zlib 0% -17% -16% -15% macOS arm64, gzip -2% -25% -28% -33% macOS arm64, zstd -14% -59% -70% -66% macOS arm64, lzma 0% -5% -7% -11% macOS arm64, bz2 -1% -1% -2% -2% Windows x64, zlib -1% +24% -32% -34% Windows x64, gzip +3% +23% -29% -31% Windows x64, zstd +1% +16% -33% -37% Windows x64, lzma +6% +3% -18% -28% Windows x64, bz2 -1% 0% -2% -3% time with this PR relative to 3.14.10 1 MiB 4 MiB 16 MiB 64 MiB macOS arm64, zlib -16% +12% 0% -1% macOS arm64, gzip -19% -1% +3% +2% macOS arm64, zstd -28% -2% -8% 0% macOS arm64, lzma -1% -2% -4% 0% macOS arm64, bz2 -2% -13% -12% -13% Windows x64, zlib -46% +2% 0% -1% Windows x64, gzip -23% +1% 0% +1% Windows x64, zstd -26% +4% +1% +2% Windows x64, lzma -17% 0% -4% +1% Windows x64, bz2 +1% +1% +1% +1% Times in ms, macOS arm64 best time 3.14.10 main this PR zlib.decompress(), 1 MiB 0.069 ms 0.058 ms 0.058 ms zlib.decompress(), 4 MiB 0.666 ms 0.899 ms 0.744 ms zlib.decompress(), 16 MiB 3.00 ms 3.58 ms 2.99 ms zlib.decompress(), 64 MiB 12.4 ms 14.5 ms 12.3 ms gzip.decompress(), 1 MiB 0.060 ms 0.050 ms 0.049 ms gzip.decompress(), 4 MiB 0.674 ms 0.885 ms 0.665 ms gzip.decompress(), 16 MiB 2.55 ms 3.68 ms 2.63 ms gzip.decompress(), 64 MiB 11.0 ms 16.6 ms 11.1 ms zstd.decompress(), 1 MiB 0.036 ms 0.030 ms 0.026 ms zstd.decompress(), 4 MiB 0.125 ms 0.295 ms 0.122 ms zstd.decompress(), 16 MiB 0.618 ms 1.87 ms 0.571 ms zstd.decompress(), 64 MiB 2.80 ms 8.17 ms 2.79 ms lzma.decompress(), 1 MiB 0.772 ms 0.761 ms 0.762 ms lzma.decompress(), 4 MiB 4.61 ms 4.80 ms 4.54 ms lzma.decompress(), 16 MiB 15.9 ms 16.3 ms 15.2 ms lzma.decompress(), 64 MiB 56.8 ms 63.6 ms 56.6 ms bz2.decompress(), 1 MiB 29.7 ms 29.4 ms 29.2 ms bz2.decompress(), 4 MiB 35.2 ms 31.0 ms 30.6 ms bz2.decompress(), 16 MiB 143 ms 127 ms 125 ms bz2.decompress(), 64 MiB 577 ms 510 ms 499 ms Times in ms, Windows x64 best time 3.14.10 main this PR zlib.decompress(), 1 MiB 0.283 ms 0.155 ms 0.154 ms zlib.decompress(), 4 MiB 1.02 ms 0.840 ms 1.04 ms zlib.decompress(), 16 MiB 4.25 ms 6.27 ms 4.27 ms zlib.decompress(), 64 MiB 17.6 ms 26.6 ms 17.5 ms gzip.decompress(), 1 MiB 0.405 ms 0.301 ms 0.310 ms gzip.decompress(), 4 MiB 1.24 ms 1.02 ms 1.25 ms gzip.decompress(), 16 MiB 4.82 ms 6.80 ms 4.81 ms gzip.decompress(), 64 MiB 19.1 ms 28.2 ms 19.4 ms zstd.decompress(), 1 MiB 0.393 ms 0.289 ms 0.291 ms zstd.decompress(), 4 MiB 1.25 ms 1.12 ms 1.30 ms zstd.decompress(), 16 MiB 4.34 ms 6.57 ms 4.40 ms zstd.decompress(), 64 MiB 16.7 ms 27.2 ms 17.1 ms lzma.decompress(), 1 MiB 0.440 ms 0.345 ms 0.366 ms lzma.decompress(), 4 MiB 6.11 ms 5.95 ms 6.14 ms lzma.decompress(), 16 MiB 10.8 ms 12.7 ms 10.4 ms lzma.decompress(), 64 MiB 26.0 ms 36.2 ms 26.1 ms bz2.decompress(), 1 MiB 30.3 ms 30.9 ms 30.6 ms bz2.decompress(), 4 MiB 35.6 ms 36.1 ms 36.0 ms bz2.decompress(), 16 MiB 145 ms 150 ms 147 ms bz2.decompress(), 64 MiB 577 ms 602 ms 585 ms Where the regression is, decompression is back at 3.14's speed: from 4 MiB on macOS and from 16 MiB on Windows, every module is within 5% of 3.14.10, except zlib at 4 MiB on macOS (+12%) and bz2, which is 12% faster than that 3.14.10 build on main and this PR alike (the build bundles its own libbz2). The gain over main is largest for zstd, whose decoder is the fastest, so the buffer was the largest share of its time; bz2's decoder is slow enough to hide the buffer either way. The decompressor objects show the same as the one-shot functions. Small outputs keep 3.15's gain: at 1 MiB this PR is within noise of main, and zlib, gzip and zstd stay well ahead of 3.14. Page faults on macOS drop from about one per 16 KiB page of output on main (0.9–1.3 in the raw runs) to none, in every module from 4 MiB. Windows at 4 MiB is 3.14's level, which is below main's. There main has grown its writer once, by a 2 MiB copy into an 8 MiB block, and writes the rest of the output straight into it; this change copies the whole output once more, into a joined result in fresh memory, because the process heap behind the UCRT's malloc returns large blocks to the OS. A finer sweep (2–24 MiB, not shown) puts the loss between the head and about 5 MiB, with a faint echo at 12 MiB, just below main's next reallocation step at 13.34 MiB; above each step main copies the writer's whole over-allocated capacity and falls behind, and no head size beats it everywhere. Against 3.14 this PR is level at every size of that sweep. On macOS there is no such crossover: the same sweep has this PR ahead at every size but one, where main's realloc() happened to land in reused memory. Compression is unchanged: the codec dominates, and zstd's buffer starts at compressBound(). Linux (glibc, default build) is unchanged: the list path is unreachable there, so the code that runs is main's. Measured anyway, as the noise calibration above. Memory, the trade-off: for outputs over 1.34 MiB, Python's peak allocation within one call goes back to 3.14's level, because the blocks and the joined result exist together at the end: 2.2–2.9× the output at 4–64 MiB, against 1.5–2.3× on main, depending on where the output ends relative to a block boundary. Measured as the process's peak commit on Windows, which tracemalloc can't see, the comparison turns around, because realloc() holds the old and the new block at once: from 4 MiB it is 2.2–3.3× the output with this PR, 2.6–4.2× on main and 3.15.2, 2.2–3.0× on 3.14.10. What that buys on macOS: main keeps about one output's worth of freed memory per call until the allocator trims, so a loop of 64 MiB decompressions grows the process by gigabytes; with this PR the footprint stays flat, as in 3.14. Growth of the process's physical footprint since before the first call, one process per case, same session as the tables above, measured with footprint.py from gh-161478 (3.15.2 behaves like main): memory the process has gained 3.14.10, any number of calls this PR, any number main, 10 calls 50 calls 100 calls 200 calls 400 calls zstd.decompress(), 1 MiB 3.0 MiB 2.1 MiB 2.6 MiB 2.6 MiB 2.6 MiB 2.6 MiB 2.6 MiB zstd.decompress(), 4 MiB 12 MiB 12 MiB 44 MiB 206 MiB 258 MiB 259 MiB 259 MiB zstd.decompress(), 16 MiB 50 MiB 48 MiB 165 MiB 600 MiB 664 MiB 664 MiB 664 MiB zstd.decompress(), 64 MiB 134 MiB 134 MiB 646 MiB 1.2 GiB 2.2 GiB 2.4 GiB 3.4 GiB zlib.decompress(), 1 MiB 2.0 MiB 1.5 MiB 1.5 MiB 1.6 MiB 1.6 MiB 1.6 MiB 1.6 MiB zlib.decompress(), 4 MiB 11 MiB 11 MiB 42 MiB 205 MiB 261 MiB 261 MiB 261 MiB zlib.decompress(), 16 MiB 46 MiB 46 MiB 163 MiB 645 MiB 979 MiB 965 MiB 920 MiB zlib.decompress(), 64 MiB 128 MiB 128 MiB 640 MiB 1.9 GiB 1.4 GiB 1.6 GiB 1.2 GiB tracemalloc's traced memory after the calls is the same on main and this PR: Python doesn't leak, the allocator keeps the memory. On Linux and Windows nothing accumulates on any of the four interpreters: glibc and the Windows heap give large blocks back, so there this PR changes only the peak. Not addressed: one-shot zstd.compress() sizes its buffer with ZSTD_compressBound() and shrinks it at the end, which this change doesn't touch. On macOS it stays 1.3× slower than 3.14 at 64 MiB; on Windows, 3.15 compresses with zstd faster than 3.14. Tests Outside @bigmemtest, the existing suites decompress past the head only in test_zstd's multithreading test (40 MiB, where libzstd supports it); nothing else in a default run reached the list-and-join path. New tests in test_zlib, test_gzip, test_bz2, test_lzma and test_zstd decompress outputs at and around the block boundaries, 1.34 MiB and 5.34 MiB, one byte below and above each: one-shot, with the decompressor object, with a max_length between the two boundaries, and (zlib) with an initial bufsize above the first one. The nine tests take under a second on a debug build and pass on main too: they cover the new code path's correctness, not its allocation pattern, which the page-fault numbers in gh-161478 show. test_zlib, test_gzip, test_bz2, test_lzma, test_zstd, test_zipfile, test_tarfile, test_shutil and test_zipimport pass on a macOS debug build, and the five compression suites pass under -R 3:3 with no leaks. On Linux (arm64, glibc), where the list path is unreachable, the five compression suites, test_zipfile and test_tarfile pass. A randomized round-trip check (a local script, not part of this PR) covers sizes at and around every block boundary: the end of the writer's four blocks and the first list blocks. It also covers one-shot and streaming APIs and random max_length and bufsize values, for zlib, gzip, bz2, lzma and zstd. It passes on the debug and PGO builds with three seeds. On Windows (x64), a debug build of this branch passes the new tests, the five compression suites, test_zipfile and test_tarfile, and the five compression suites under -R 3:3 with no leaks; the round-trip check passes on its debug and PGO builds with three seeds. Backport: I'd suggest backporting to 3.15, since the slowdown and the footprint growth are in 3.15.2 and the change only restores what 3.14 did, with no API change. I used an AI assistant to investigate this and draft the patch; I reviewed and tested the changes myself.

Open Graph Description: Fixes gh-161478: decompression into large outputs got slower in 3.15 on macOS and Windows, from a few MiB of output (4 on macOS, 16 on Windows), after f262297 (gh-139877) moved _BlocksOutputBuffer ...

X Description: Fixes gh-161478: decompression into large outputs got slower in 3.15 on macOS and Windows, from a few MiB of output (4 on macOS, 16 on Windows), after f262297 (gh-139877) moved _BlocksOutputBuffer ...

Opengraph URL: https://github.com/python/cpython/pull/161482

X: @github

direct link

Domain: github.com

react-profiling0
react-import-mapreact,react-dom,react-dom/client,react-dom/profiling,react-is,react/compiler-runtime,react/jsx-dev-runtime,react/jsx-runtime,scheduler@b8488ee2c322
route-pattern/:user_id/:repository/pull/:id/files(.:format)
route-controllerpull_requests
route-actionfiles
fetch-noncev2:3df93a8f-8ecc-2ebc-42ca-ca218c127dbd
current-catalog-service-hashae870bc5e265a340912cde392f23dad3671a0a881730ffdadd82f2f57d81641b
request-id9582:76F02:3EE4738:505A98C:6B0216BD
html-safe-nonce98a1a532845a783128788287e25dc30d07d61206a26dff618d8e7e7d7050bab1
visitor-payloadeyJyZWZlcnJlciI6IiIsInJlcXVlc3RfaWQiOiI5NTgyOjc2RjAyOjNFRTQ3Mzg6NTA1QTk4Qzo2QjAyMTZCRCIsInZpc2l0b3JfaWQiOiIzOTg2MzkxNDM3MjAxOTU3NzMiLCJyZWdpb25fZWRnZSI6ImlhZCIsInJlZ2lvbl9yZW5kZXIiOiJpYWQifQ==
visitor-hmac288c44d8a9329548fa6d1f50cbe1c145f66b407d8dbe312c1dbeb68adb89ecd9
hovercard-subject-tagpull_request:5187073863
github-keyboard-shortcutsrepository,pull-request-list,pull-request-conversation,pull-request-files-changed,copilot
google-site-verificationApib7-x98H0j5cPqHWwSMm6dNU4GmODRoqxLiDzdx9I
octolytics-urlhttps://collector.github.com/github/collect
analytics-location///pull_requests/show/files
fb:app_id1401488693436528
apple-itunes-appapp-id=1477376905, app-argument=https://github.com/python/cpython/pull/161482/files
twitter:imagehttps://avatars.githubusercontent.com/u/4232812?s=400&v=4
twitter:cardsummary_large_image
og:imagehttps://avatars.githubusercontent.com/u/4232812?s=400&v=4
og:image:altFixes gh-161478: decompression into large outputs got slower in 3.15 on macOS and Windows, from a few MiB of output (4 on macOS, 16 on Windows), after f262297 (gh-139877) moved _BlocksOutputBuffer ...
og:site_nameGitHub
og:typeobject
hostnamegithub.com
expected-hostnamegithub.com
None87e774ee7c6b3bee0490394e4d3e57ec9dbc956f74b87ea685dc8996c1676456
turbo-cache-controlno-preview
diff-viewunified
go-importgithub.com/python/cpython git https://github.com/python/cpython.git
octolytics-dimension-user_id1525981
octolytics-dimension-user_loginpython
octolytics-dimension-repository_id81598961
octolytics-dimension-repository_nwopython/cpython
octolytics-dimension-repository_publictrue
octolytics-dimension-repository_is_forkfalse
octolytics-dimension-repository_network_root_id81598961
octolytics-dimension-repository_network_root_nwopython/cpython
turbo-body-classeslogged-out env-production page-responsive full-width
disable-turbotrue
browser-stats-urlhttps://api.github.com/_private/browser/stats
browser-errors-urlhttps://api.github.com/_private/browser/errors
release423dbb54c5e7bfecc838eb387c1435eb3d0e760d
ui-targetfull
theme-color#1e2327
color-schemelight dark

Links:

Skip to contenthttps://github.com/python/cpython/pull/161482/files#start-of-content
https://github.com/
Sign inhttps://github.com/login?return_to=https%3A%2F%2Fgithub.com%2Fpython%2Fcpython%2Fpull%2F161482%2Ffiles
GitHubRepository hosting, code review, and collaborationhttps://github.com/platform
CopilotAI coding agent for tasks, projects, and automated workhttps://github.com/features/copilot
ActionsCI/CD and automationhttps://github.com/features/actions
Code QualityQuality analysis, fixes, and enforcementhttps://github.com/features/code-quality
CodespacesCloud development environmentshttps://github.com/features/codespaces
Code SecurityVulnerability detection and remediationhttps://github.com/security/advanced-security/code-security
Secret ProtectionSecret scanning and leak preventionhttps://github.com/security/advanced-security/secret-protection
All Featureshttps://github.com/features
Documentationhttps://docs.github.com/en
Traininghttps://learn.github.com/skills
Certificationshttps://learn.github.com/credentials
Build with Copilothttps://github.com/features/ai
Customize Copilothttps://docs.github.com/en/copilot/how-tos/copilot-on-github/customize-copilot/customize-copilot-overview
Integrationshttps://github.com/marketplace
Trending reposhttps://github.com/trending
GitHub Sponsorshttps://github.com/open-source/sponsors
Developer articleshttps://github.com/resources/articles
ChangelogYour source for the latest features, improvements, and fixes across GitHub products, all in one place.https://github.blog/changelog/
Enterpriseshttps://github.com/enterprise
Teamhttps://github.com/team
Startupshttps://github.com/enterprise/startups
Nonprofitshttps://github.com/solutions/industry/nonprofits
Healthcarehttps://github.com/solutions/industry/healthcare
Financial serviceshttps://github.com/solutions/industry/financial-services
Manufacturinghttps://github.com/solutions/industry/manufacturing
Governmenthttps://github.com/solutions/industry/government
See all industrieshttps://github.com/solutions/industry
Why GitHubhttps://github.com/why-github
Customer Storieshttps://github.com/customer-stories
Business insightshttps://github.com/solutions/executive-insights
Trust Centerhttps://github.com/trust-center
GitHub Advanced SecurityEnterprise-grade code securityhttps://github.com/security/advanced-security
Copilot for businessEnterprise-grade AI featureshttps://github.com/features/copilot/copilot-business
Premium SupportEnterprise-grade 24/7 supporthttps://github.com/enterprise/premium-support
Documentationhttps://docs.github.com/en
Resource centerhttps://github.com/resources
Bloghttps://github.blog/
Events & webinarshttps://github.com/resources/events
Community Forumhttps://github.com/orgs/community/discussions
Customer Storieshttps://github.com/customer-stories
Integrationshttps://github.com/marketplace
Customer supporthttps://help.github.com/
Traininghttps://learn.github.com/skills
GitHub Statushttps://www.githubstatus.com/
Abouthttps://github.com/about
Partnershttps://github.com/partners
Changeloghttps://github.blog/changelog/
GitHub UniverseThe event for developers, by developers — talks, workshops, and what’s next for GitHub.https://githubuniverse.com/
Pricinghttps://github.com/pricing
Sign inhttps://github.com/login?return_to=https%3A%2F%2Fgithub.com%2Fpython%2Fcpython%2Fpull%2F161482%2Ffiles
Sign uphttps://github.com/signup?ref_cta=Sign+up&ref_loc=header+logged+out&ref_page=%2F%3Cuser-name%3E%2F%3Crepo-name%3E%2Fpull_requests%2Fshow%2Ffiles&source=header-repo&source_repo=python%2Fcpython
Reloadhttps://github.com/python/cpython/pull/161482/files
Reloadhttps://github.com/python/cpython/pull/161482/files
Reloadhttps://github.com/python/cpython/pull/161482/files
pythonhttps://github.com/python
cpythonhttps://github.com/python/cpython
Codehttps://github.com/python/cpython
Issues5k+ (5k+)https://github.com/python/cpython/issues
Pull requests2.4k (2.4k)https://github.com/python/cpython/pulls
Actionshttps://github.com/python/cpython/actions
Projectshttps://github.com/python/cpython/projects
Security and qualityhttps://github.com/python/cpython/security
Insightshttps://github.com/python/cpython/insights
Sign up for GitHub https://github.com/signup?return_to=%2Fpython%2Fcpython%2Fissues%2Fnew%2Fchoose
terms of servicehttps://docs.github.com/terms
privacy statementhttps://docs.github.com/privacy
Sign inhttps://github.com/login?return_to=%2Fpython%2Fcpython%2Fissues%2Fnew%2Fchoose
maikschulzehttps://github.com/maikschulze
python:mainhttps://github.com/python/cpython/tree/main
maikschulze:compression-output-bufferhttps://github.com/maikschulze/cpython/tree/compression-output-buffer
Conversation 1 https://github.com/python/cpython/pull/161482
Commits 1 https://github.com/python/cpython/pull/161482/commits
Checks 54 https://github.com/python/cpython/pull/161482/checks
Files changed https://github.com/python/cpython/pull/161482/files
Please reload this pagehttps://github.com/python/cpython/pull/161482/files
gh-161478: Collect large compression outputs in blocks again, except with glibc https://github.com/python/cpython/pull/161482/files#top
Show all changes 1 commit https://github.com/python/cpython/pull/161482/files
cceb782 gh-161478: Collect large compression outputs in blocks again, except … maikschulze Nov 6, 2026 https://github.com/python/cpython/pull/161482/commits/cceb7826fc31848a4675533db497583d0ca4a825
Clear filters https://github.com/python/cpython/pull/161482/files
Please reload this pagehttps://github.com/python/cpython/pull/161482/files
Please reload this pagehttps://github.com/python/cpython/pull/161482/files
pycore_blocks_output_buffer.h https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
test_bz2.py https://github.com/python/cpython/pull/161482/files#diff-25018b1e1b3efce8819bad84e95a0f090c29dabfd095639626b5dcfd04ed3d20
test_gzip.py https://github.com/python/cpython/pull/161482/files#diff-9929cddaa6d51e36fc67f9426bd70f73aa995b01f4ce0a5d9b93c83a3a5c4587
test_lzma.py https://github.com/python/cpython/pull/161482/files#diff-c0d354f43bb17a1ab769897c04ad70a9d5364b876cde66a0fa6ebf50f7db1fb6
test_zlib.py https://github.com/python/cpython/pull/161482/files#diff-029a0ce1b89a517af9cd486ab867641e1c90e4ac089dded48ee1b65e604b64aa
test_zstd.py https://github.com/python/cpython/pull/161482/files#diff-d347ebfcbcfd73ae33135b08623594cc45342d42b774fbbc80e508176f860d5d
2026-11-05-21-00-00.gh-issue-161478.OutBuf.rst https://github.com/python/cpython/pull/161482/files#diff-18cbe43971eb059edab048241fa3663174e8c62232de90a664fdc1853e38315d
Include/internal/pycore_blocks_output_buffer.hhttps://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
View file https://github.com/python/cpython/blob/cceb7826fc31848a4675533db497583d0ca4a825/Include/internal/pycore_blocks_output_buffer.h
Open in desktop https://desktop.github.com
https://github.co/hiddenchars
https://github.com/python/cpython/pull/161482/{{ revealButtonHref }}
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
https://github.com/python/cpython/pull/161482/files#diff-137d424d36904214cf556dd06f59386410f53a13f222a46ceeed30b7a9f6c234
Lib/test/test_bz2.pyhttps://github.com/python/cpython/pull/161482/files#diff-25018b1e1b3efce8819bad84e95a0f090c29dabfd095639626b5dcfd04ed3d20
View file https://github.com/python/cpython/blob/cceb7826fc31848a4675533db497583d0ca4a825/Lib/test/test_bz2.py
Open in desktop https://desktop.github.com
https://github.co/hiddenchars
https://github.com/python/cpython/pull/161482/{{ revealButtonHref }}
https://github.com/python/cpython/pull/161482/files#diff-25018b1e1b3efce8819bad84e95a0f090c29dabfd095639626b5dcfd04ed3d20
https://github.com/python/cpython/pull/161482/files#diff-25018b1e1b3efce8819bad84e95a0f090c29dabfd095639626b5dcfd04ed3d20
Lib/test/test_gzip.pyhttps://github.com/python/cpython/pull/161482/files#diff-9929cddaa6d51e36fc67f9426bd70f73aa995b01f4ce0a5d9b93c83a3a5c4587
View file https://github.com/python/cpython/blob/cceb7826fc31848a4675533db497583d0ca4a825/Lib/test/test_gzip.py
Open in desktop https://desktop.github.com
https://github.co/hiddenchars
https://github.com/python/cpython/pull/161482/{{ revealButtonHref }}
https://github.com/python/cpython/pull/161482/files#diff-9929cddaa6d51e36fc67f9426bd70f73aa995b01f4ce0a5d9b93c83a3a5c4587
Lib/test/test_lzma.pyhttps://github.com/python/cpython/pull/161482/files#diff-c0d354f43bb17a1ab769897c04ad70a9d5364b876cde66a0fa6ebf50f7db1fb6
View file https://github.com/python/cpython/blob/cceb7826fc31848a4675533db497583d0ca4a825/Lib/test/test_lzma.py
Open in desktop https://desktop.github.com
https://github.co/hiddenchars
https://github.com/python/cpython/pull/161482/{{ revealButtonHref }}
https://github.com/python/cpython/pull/161482/files#diff-c0d354f43bb17a1ab769897c04ad70a9d5364b876cde66a0fa6ebf50f7db1fb6
Please reload this pagehttps://github.com/python/cpython/pull/161482/files
Please reload this pagehttps://github.com/python/cpython/pull/161482/files
https://github.com
Termshttps://docs.github.com/site-policy/github-terms/github-terms-of-service
Privacyhttps://docs.github.com/site-policy/privacy-policies/github-privacy-statement
Securityhttps://github.com/security
Statushttps://www.githubstatus.com/
Communityhttps://github.community/
Docshttps://docs.github.com/
Contacthttps://support.github.com?tags=dotcom-footer

Viewport: width=device-width


URLs of crawlers that visited me.