Use `cp -c` when copying files #17373

tesaguri · 2024-05-27T03:37:01Z

Have you followed the guidelines in our Contributing document?
Have you checked to ensure there aren't other open Pull Requests for the same change?
Have you added an explanation of what your changes do and why you'd like us to include them?
Have you written new tests for your changes? Here's an example.
Have you successfully run brew style with your changes locally?
Have you successfully run brew typecheck with your changes locally?
Have you successfully run brew tests with your changes locally?

This patch makes Homebrew use cp --reflink=auto if coreutils formula is installed, which significantly speeds up build/installation operations and reduces the risk of exhausting the storage during operations especially when dealing with large formulae.

I will work on new tests if the idea sounds good to you.

This module determines the `cp` command to use based on availability of the `coreutils` formula and optimizes the command invocation to prefer a lightweight copy-on-write clone, which is significantly faster than a full file copy and helps to reduce the risk of exhausting the storage during the operation.

This replaces `FileUtils.cp` and `system_command! "cp"` with the new `Utils::Cp` utility where it is expected that the performance improvement outweighs the cost of the system command invocation.

Bo98 · 2024-05-27T03:47:06Z

There is a cost of shelling out that should be considered here too. FileUtils.cp does not shell out. You should present your case with numbers for each of the call sites changed with both small and large payloads.

FWIW what happens under the hood with reflink is it calls clonefile on macOS and does a ioctl(FICLONE, ...) on Linux, though I guess there isn't Ruby bindings for those.

macOS's default cp supports a similar feature with the -c flag since I think macOS 10.13, so that seems like a much easier option for the cask code that already shells out.

tesaguri · 2024-05-27T04:02:49Z

There is a cost of shelling out that should be considered here too. FileUtils.cp does not shell out. You should present your case with numbers for each of the call sites changed with both small and large payloads.

I have tried to limit the use of the system command to places that I guessed to be likely to deal with large payloads, but yes, that's only wild guess and I surely need some benchmarks.

macOS's default cp supports a similar feature with the -c flag since I think macOS 10.13, so that seems like a much easier option for the cask code that already shells out.

I have considered that too, but I didn't adopt it because it fails if the clonefile syscall isn't applicable to the source and destination of the copy, as I wrote in https://github.com/Homebrew/brew/pull/17373/files#diff-52b1a61a5f4b7c031445c5f76c206db7c60089cb7204d7eeacbef0a0924e8553R61-R64.

(And sorry for the test failure. I forgot to run the test after rebasing on latest master. I'll fix it.)

Bo98 · 2024-05-27T04:11:49Z

have considered that too, but I didn't adopt it because it fails if the clonefile syscall isn't applicable to the source and destination of the copy, as I wrote in https://github.com/Homebrew/brew/pull/17373/files#diff-52b1a61a5f4b7c031445c5f76c206db7c60089cb7204d7eeacbef0a0924e8553R61-R64.

Ok fair point, though it does also seem macOS 14 shipped with a change to -c:

     -c    copy files using clonefile(2).  Note that if clonefile(2) is not
           supported for the target filesystem, then cp will fallback to using
           copyfile(2) instead to ensure the copy still succeeds.

Under the hood, it does this by checking if clonefile returns ENOTSUP. Are you still seeing the behaviour you describe on macOS 14?

tesaguri · 2024-05-27T04:38:40Z

Are you still seeing the behaviour you describe on macOS 14?

No. I was just writing in terms of the compatibility with old systems and it's not that I observe the behavior on the latest system.

Perhaps I can make the module change the strategy based on macOS version.

Bo98 · 2024-05-27T05:33:10Z

Even gating the whole feature to >= macOS 14 is probably ok. It would likely significantly simplify the code and that seems acceptable given we've gone on fine without the feature this long and older macOS versions cover only ~15% of users, and that figure will only decrease further (macOS 15 is announced next month).

MikeMcQuaid · 2024-05-27T07:05:53Z

Thanks for the PR @tesaguri!

Even gating the whole feature to >= macOS 14 is probably ok.

Yes, this seems like the right approach.

You should present your case with numbers for each of the call sites changed with both small and large payloads.

Not just microbenchmarks either but the benchmarks of actually installing e.g. a relevant formula/cache.

Do this before adding/changing any tests as if the performance gain is not significant this won't be accepted.

As per review feedback: Homebrew#17373 (comment)

MikeMcQuaid

Thanks @tesaguri, looking great and we're almost there! ❤️

Library/Homebrew/utils/cp.rb

Library/Homebrew/dev-cmd/bottle.rb

As per review feedback: Homebrew#17373 (review)

tesaguri · 2024-06-07T13:00:25Z

I've done a quick benchmark of the execution time of brew [un]install with a cask (.dmg) and a bottled formula (.tar.gz).

The following tables are output of the hyperfine benchmarking tool and the Command names master-* and reflink-* represent the results of Homebrew:master branch and tesaguri:cp-reflink branch (this patch) respectively.

brew install libreoffice:

Command	Mean [s]	Min [s]	Max [s]	Relative
`master-install-libreoffice`	84.196 ± 6.290	78.934	100.503	1.41 ± 0.13
`reflink-install-libreoffice`	59.804 ± 2.992	57.261	67.991	1.00

brew uninstall libreoffice:

Command	Mean [s]	Min [s]	Max [s]	Relative
`master-uninstall-libreoffice`	45.217 ± 2.183	41.470	49.262	3.17 ± 0.24
`reflink-uninstall-libreoffice`	14.274 ± 0.844	12.609	15.150	1.00

brew install llvm:

Command	Mean [s]	Min [s]	Max [s]	Relative
`master-install-llvm`	137.381 ± 4.537	132.001	148.779	1.29 ± 0.05
`reflink-install-llvm`	106.714 ± 1.671	104.831	109.686	1.00

brew uninstall llvm:

Command	Mean [s]	Min [s]	Max [s]	Relative
`master-uninstall-llvm`	9.514 ± 0.782	8.597	10.871	1.00
`reflink-uninstall-llvm`	9.684 ± 0.846	8.459	11.047	1.02 ± 0.12

(The uninstallation seems to be unaffected by the change, as expected.)

Also, I've tested with the bottle file as the argument just to be sure.

brew install "$(brew --cache llvm)":

Command	Mean [s]	Min [s]	Max [s]	Relative
`master-install-llvm_bottle`	148.959 ± 2.346	146.004	152.294	1.15 ± 0.02
`reflink-install-llvm_bottle`	129.275 ± 1.413	127.639	132.201	1.00

So far, the results are easily expected given the large size of these cask and formula (and there are no additional command invocation cost for casks anyway).

But the pessimistic case would be small formulae for which the cost of system command invoctions relative to the IO cost is expected to be higher.

To simulate the most pessimistic case, I've built a bottle with only a single zero-sized file and measured the installation time of it (the source of the formula will be shown later).

Because the bottle is so small that it is hard to calibrate the time taken by unrelated tasks for sure, I've created 100 such bottles and installed them at once in each brew command invocation.

brew install -- test-formula-tiny*--0.0.0.all.bottle.tar.gz:

Command	Mean [s]	Min [s]	Max [s]	Relative
`master-install-test-formula-tiny000_bottle.test-formula-tiny001_bottle.test-formula-tiny002_bottle`	31.911 ± 0.579	31.014	33.254	1.21 ± 0.03
`reflink-install-test-formula-tiny000_bottle.test-formula-tiny001_bottle.test-formula-tiny002_bottle`	26.359 ± 0.294	25.876	26.873	1.00

brew uninstall -- test-formula-tiny*--0.0.0.all.bottle.tar.gz:

Command	Mean [s]	Min [s]	Max [s]	Relative
`master-uninstall-test-formula-tiny000_bottle.test-formula-tiny001_bottle.test-formula-tiny002_bottle`	18.868 ± 0.475	18.263	19.584	1.80 ± 0.06
`reflink-uninstall-test-formula-tiny000_bottle.test-formula-tiny001_bottle.test-formula-tiny002_bottle`	10.505 ± 0.258	10.099	10.975	1.00

The improvement in the uninstallation time is surprising to me. I think I'm missing something… Perhaps, this has something to do with the direct uninstallation from bottle files, which involves extraction of the bottles, I guess?

Anyway, it seems that the change improves the performance even in the pessimistic case (at least it doesn't degrade the performance).

(I haven't come up with test cases of other download strategies like Xz that represent realistic use cases well enough. Also, I couldn't figure out any code paths that lead to the cp invocation in Pathname#cp_sub_path method, because it only calls mkpath if self.directory? is true and all of its callers seem to call the method only when directory? is true.)

Reproduction

Source code of the test formula:

class TestFormulaTiny000 < Formula
  url "file:///dev/null"
  version "0.0.0"
  sha256 "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"

  def install
    mkdir_p "test-tiny"
    mv "null", "test-tiny/000"
    share.install "test-tiny"
  end
end

Duplicate it and build them with the following command:

$ for i in {001..099}; do sed "s/000/$i/g" test-formula-tiny000.rb > "test-formula-tiny$i.rb"; done
$ brew install --build-bottle -- test-formula-tiny{000..099}

Here is a helper script for the benchmark (requires hyperfine formula to be installed):

bench.sh

#!/bin/bash

set -e

# WORKTREE=(path to the worktree of the fork)
WORKTREE_BREW="$WORKTREE/bin/brew"

HYPERFINE_FLAGS='--warmup=3'
BREW_INSTALL_FLAGS='--ignore-dependencies --skip-cask-deps'
BREW_UNINSTALL_FLAGS='--ignore-dependencies'

cd "$(dirname "$0")"

subcommand="$1"
shift

args=()
pkgs=()
for arg in "$@"; do
	args+=("$(printf '%q' "$arg")")
	if ! [[ "$arg" = -* ]]; then
		if [[ "$arg" = *.bottle.tar.gz ]]; then
			arg="${arg##*/}"
			arg="${arg%.bottle.tar.gz}"
			arg="${arg%--*}"
			arg="${arg#*--}_bottle"
		fi
		pkgs+=("$arg")
	fi
done
pkgs_label="$(IFS='.'; echo "${pkgs[*]:0:3}")"
pkgs_label="${pkgs_label:0:240}"

for env in $(export | grep '^HOMEBREW_'); do
	unset "$env"
done

export HOMEBREW_CURL_RETRIES=0
export HOMEBREW_NO_AUTOREMOVE=1
export HOMEBREW_NO_AUTO_UPDATE=1
export HOMEBREW_NO_INSTALLED_DEPENDENTS_CHECK=1
export HOMEBREW_NO_INSTALL_CLEANUP=1
export HOMEBREW_NO_INSTALL_UPGRADE=1

case "$subcommand" in
install)
	fetch=()
	for arg in "${args[@]}"; do
		case "$arg" in
		-*|*.bottle.tar.gz)
			;;
		*)
			fetch+=("$arg")
			;;
		esac
	done
	mkdir -p results
	hyperfine $HYPERFINE_FLAGS --shell=none \
		--export-json="results/install-$pkgs_label.json" \
		--export-markdown="results/install-$pkgs_label.md" \
		${fetch:+--setup="sh -c 'brew fetch ${fetch[*]} || true'"} \
		--command-name="master-install-$pkgs_label" \
		--prepare="brew uninstall --force $BREW_UNINSTALL_FLAGS ${args[*]}" \
		"brew install $BREW_INSTALL_FLAGS ${args[*]}" \
		--command-name="reflink-install-$pkgs_label" \
		--prepare="$(printf '%q' "$WORKTREE_BREW") uninstall --force $BREW_UNINSTALL_FLAGS ${args[*]}" \
		"$(printf '%q' "$WORKTREE_BREW") install $BREW_INSTALL_FLAGS ${args[*]}"
	;;
uninstall)
	mkdir -p results
	hyperfine $HYPERFINE_FLAGS --shell=none \
		--export-json="results/uninstall-$pkgs_label.json" \
		--export-markdown="results/uninstall-$pkgs_label.md" \
		--command-name="master-uninstall-$pkgs_label" \
		--prepare="brew install --force $BREW_INSTALL_FLAGS ${args[*]}" \
		"brew uninstall $BREW_UNINSTALL_FLAGS ${args[*]}" \
		--command-name="reflink-uninstall-$pkgs_label" \
		--prepare="$(printf '%q' "$WORKTREE_BREW") install --force $BREW_INSTALL_FLAGS ${args[*]}" \
		"$(printf '%q' "$WORKTREE_BREW") uninstall $BREW_UNINSTALL_FLAGS ${args[*]}"
	;;
*)
	echo "Usage: ${0##*/} install|uninstall PACKAGE..."
	exit 1
	;;
esac

Run the benchmark with the following commands:

$ brew fetch -- libreoffice llvm
$ : Here, I have disconnected my machine from the internet to make `curl` fail immidiately,
$ : reducing the variance due to the network condition.
$ export WORKTREE=(path to the worktree of the fork); \
	caffeinate ./bench.sh install -- libreoffice; \
	caffeinate ./bench.sh uninstall -- libreoffice; \
	caffeinate ./bench.sh install -- llvm; \
	caffeinate ./bench.sh uninstall -- llvm; \
	caffeinate ./bench.sh install -- "$(brew --cache -- llvm)"; \
	caffeinate ./bench.sh install -- test-formula-tiny{000..099}--0.0.0.all.bottle.tar.gz; \
	caffeinate ./bench.sh uninstall -- test-formula-tiny{000..099}--0.0.0.all.bottle.tar.gz

MikeMcQuaid

Some more naming thoughts. This is very nearly ready to 🚢! Very pleased by performance increases!

Library/Homebrew/utils/cp.rb

MikeMcQuaid · 2024-06-07T13:16:50Z

Thanks for the detailed write-up @tesaguri! Are you using hyperfine for these benchmarks? If so, use the --warmup option if you're not already as it may help explain some of the unexpected results.

As per review feedback: Homebrew#17373 (review) Co-authored-by: Mike McQuaid <[email protected]>

Homebrew#17373 (review) Co-authored-by: Mike McQuaid <[email protected]>

tesaguri · 2024-06-08T01:29:28Z

Are you using hyperfine for these benchmarks? If so, use the --warmup option if you're not already as it may help explain some of the unexpected results.

Yes, I'm already using --warmup=3.

Library/Homebrew/utils/cp.rb

Bo98 · 2024-06-08T04:32:59Z

This is pretty cool and has inspired me to look into adding clonefile to Ruby upstream. The Linux equivalent appears to already be implemented so seems reasonable to bring macOS in line.

`FileUtils.cp` is implemented with the lightweight `copy_file_range(2)` syscall on Linux, so it's more performant than the plain `cp` command on that platform. cf. Homebrew#17373 (review)

tesaguri · 2024-06-08T12:32:30Z

https://github.com/Homebrew/brew/actions/runs/9428441435/job/25973762028?pr=17373#step:12:42

Well, I have forgotten that UnpackStrategy::Directory is also using the cp command.

https://github.com/Homebrew/brew/blob/master/Library/Homebrew/unpack_strategy/directory.rb#L23-L26

Reverting this code to master should fix the test, but I think that UnpackStrategy::Directory is one of the most performance-critical parts and that simply reverting this change spoils the point of the patch.

Perhaps, we can add yet another utility method (Edit: or a new keyword argument for existing methods?) that is guaranteed to use cp command, though I expect that would increase code complexity.

Another option is to revert the latest commit (9156891), but that would lead to a significant performance regression on Linux.

Or should we abandon the whole patch and wait for Ruby upstream to use clonefile (if ever)? But UnpackStrategy::Directory and Cask::Artifact::Moved (which uses cp command) would be stuck in the status-quo performance that way.

This fixes the test for `UnpackStrategy::Directory`, which needs the `cp` command.

tesaguri · 2024-06-09T01:45:56Z

Added force_command (the name is up for bikeshedding) keyword arguments to force the use of the shell command in 9156891, which I think ended up reasonably simple (compared to the original idea of adding dedicated methods instead).

Bo98 · 2024-06-09T04:45:28Z

I gave it a go: Bo98/ruby@6fc57a2.

$ mkfile -n 1g big

$ hyperfine --warmup 3 --prepare "rm big2 || true" 'ruby-orig -e "IO.copy_stream(\"big\", \"big2\")"'
Benchmark 1: ruby-orig -e "IO.copy_stream(\"big\", \"big2\")"
  Time (mean ± σ):     251.1 ms ±   8.3 ms    [User: 23.1 ms, System: 141.8 ms]
  Range (min … max):   238.6 ms … 264.5 ms    11 runs

$ hyperfine --warmup 3 --prepare "rm big2 || true" 'ruby-new -e "IO.copy_stream(\"big\", \"big2\")"'
Benchmark 1: ruby-new -e "IO.copy_stream(\"big\", \"big2\")"
  Time (mean ± σ):      25.9 ms ±   0.9 ms    [User: 21.5 ms, System: 3.6 ms]
  Range (min … max):    25.2 ms …  33.7 ms    97 runs

Works on macOS 10.13+.
I'll discuss this with upstream. If it is accepted, I will backport this to our Ruby. One catch is that in order for it to take effect on Intel, we would need to bump the minimum macOS to 10.12 (arm64 already has a macOS 11 minimum so that's ok). There's graceful fallback on 10.12, but not for 10.11. This doesn't necessarily need to happen immediately - we can backport and the configure checks will automatically only apply the new code to arm64 until we bump the minimum.

Or should we abandon the whole patch and wait for Ruby upstream to use clonefile (if ever)?

I am ok with this PR as long as the commits are structured in a way that it's easy to revert the unpack_strategy changes if Ruby itself gets patched.

Casks will still use shelling out due to sudo needs, and those are also macOS only.

Bo98 · 2024-06-09T04:53:57Z

Well, I have forgotten that UnpackStrategy::Directory is also using the cp command.

Hmm that's annoying. Really it should use FileUtils, but it seems it doesn't for some reason. Will check and see if that reason is still valid.

MikeMcQuaid

One more question/discussion on some names and I think we're good to go here, thanks @tesaguri!

Library/Homebrew/utils/cp.rb

As per review feedback: Homebrew#17373 (comment)

tesaguri · 2024-06-12T03:56:42Z

I've added tests for the new module in 4edbbfd.

I'm going to edit the commit history after the review completes, so that it's easier to partially revert the change if/when FileUtils supports clonefile(2).

MikeMcQuaid

Great work @tesaguri! One more comment and then I'm ✅, thanks again for such great work here!

Library/Homebrew/utils/copy.rb

MikeMcQuaid · 2024-06-13T07:57:07Z

Thanks so much for your first contribution! Without people like you submitting PRs we couldn't run this project. You rock, @tesaguri; this is one of the highest quality first PRs I've ever seen in 15 years maintaining Homebrew! I hope to see many more from you ❤️

apainintheneck · 2024-06-18T04:24:04Z

It looks like this change is causing some problems for folks so I opened a revert PR.

Revert "Use cp -c when copying files" #17526

tesaguri · 2024-06-18T14:00:23Z

Ah, sorry for the regression. Is the revert intended to be temporary?

If so, it looks like the issues could be fixed by using fully-qualified path for the cp command, i.e., "/bin/cp", though I haven't closely looked into the issues yet and it would definitely need some sort of regression tests.

apainintheneck · 2024-06-18T14:40:35Z

Ah, sorry for the regression. Is the revert intended to be temporary?

No worries. It happens. We're usually open to retrying PRs that get reverted as long as the original issues get addressed. I'm just not super familiar with this code personally so I didn't see an obvious way to fix forward last night. Feel free to open another PR for that.

MikeMcQuaid · 2024-06-19T11:52:57Z

it looks like the issues could be fixed by using fully-qualified path for the cp command, i.e., "/bin/cp",

This sounds great, thanks!

jmcbsimmonds · 2024-06-22T09:21:53Z

Hello,

It seems adding "-c" option to cp can cause an error on MacOS Sonoma 14.9:

For example:
Error: zoom: Failure while executing; /usr/bin/env cp -p -c /Volumes/Home/Users/username/Library/Caches/Homebrew/downloads/960197a91564aadc18ea997951366349d9a9604c79f350db9b259cdc9000f7e2--zoomusInstallerFull.pkg /usr/local/Caskroom/zoom/6.1.0.35886/zoomusInstallerFull.pkg exited with 1. Here's the output:
cp: /usr/local/Caskroom/zoom/6.1.0.35886/zoomusInstallerFull.pkg: clonefile failed: Cross-device link

The / filesystem is APFS, the /Volumes/Home filesystem is APFS (Case Sensitive)

Same error with Microsoft-Edge.

HOMEBREW_VERSION: 4.3.6
ORIGIN: https://github.com/Homebrew/brew
HEAD: e8430b2
Last commit: 6 days ago
Core tap JSON: 22 Jun 05:42 UTC
Core cask tap JSON: 22 Jun 05:42 UTC
HOMEBREW_PREFIX: /usr/local
HOMEBREW_CASK_OPTS: []
HOMEBREW_MAKE_JOBS: 16
Homebrew Ruby: 3.3.3 => /usr/local/Homebrew/Library/Homebrew/vendor/portable-ruby/3.3.3/bin/ruby
CPU: 16-core 64-bit kabylake
Clang: 15.0.0 build 1500
Git: 2.45.2 => /usr/local/bin/git
Curl: 8.6.0 => /usr/bin/curl
macOS: 14.5-x86_64
CLT: 15.3.0.0.1.1708646388
Xcode: 15.4

Is there any config option to turn off -c option?

Kind regards.

apainintheneck · 2024-06-22T17:17:51Z

@jmcbsimmonds This change has been reverted but has not made it into the most recent release yet. The weekly release usually comes out Monday-ish.

MikeMcQuaid · 2024-06-24T12:02:41Z

This is released/fixed in Homebrew 4.3.7 now: https://github.com/Homebrew/brew/releases/tag/4.3.7

tesaguri added 2 commits May 27, 2024 12:11

Use Utils::Cp to copy files

deaac7c

This replaces `FileUtils.cp` and `system_command! "cp"` with the new `Utils::Cp` utility where it is expected that the performance improvement outweighs the cost of the system command invocation.

tesaguri changed the title ~~Cp reflink~~ Use cp --reflink=auto when copying files May 27, 2024

tesaguri added 4 commits May 27, 2024 17:48

Utils::Cp: Fix tests

b8af1c5

Utils::Cp: Use cp from macOS

942906b

Utils::Cp: Drop special case for coreutils cp

b4dcb94

As per review feedback: Homebrew#17373 (comment)

Utils::Cp: Fix Linux tests

a5500aa

tesaguri changed the title ~~Use cp --reflink=auto when copying files~~ Use cp -c when copying files Jun 6, 2024

MikeMcQuaid reviewed Jun 7, 2024

View reviewed changes

Library/Homebrew/utils/cp.rb Outdated Show resolved Hide resolved

Library/Homebrew/utils/cp.rb Outdated Show resolved Hide resolved

Library/Homebrew/utils/cp.rb Outdated Show resolved Hide resolved

Library/Homebrew/dev-cmd/bottle.rb Outdated Show resolved Hide resolved

tesaguri added 3 commits June 7, 2024 19:03

Utils::Cp: Rename copy* methods to copy*_with_attributes

5885210

As per review feedback: Homebrew#17373 (review)

Utils::Cp: Move macOS-specific code to extend/os/mac

7cfcc59

Utils::Cp: Revert the use of Utils::Cp in dev-cmd

d9239fa

As per review feedback: Homebrew#17373 (review)

MikeMcQuaid reviewed Jun 7, 2024

View reviewed changes

Library/Homebrew/utils/cp.rb Outdated Show resolved Hide resolved

Library/Homebrew/utils/cp.rb Outdated Show resolved Hide resolved

tesaguri and others added 2 commits June 8, 2024 07:58

Utils::Cp: Remove copy prefix from method name

b2ddeec

As per review feedback: Homebrew#17373 (review) Co-authored-by: Mike McQuaid <[email protected]>

Apply suggestions from code review

a4271fd

Homebrew#17373 (review) Co-authored-by: Mike McQuaid <[email protected]>

Bo98 reviewed Jun 8, 2024

View reviewed changes

Library/Homebrew/utils/cp.rb Outdated Show resolved Hide resolved

Utils::Cp: Use FileUtils.cp on Linux

9156891

`FileUtils.cp` is implemented with the lightweight `copy_file_range(2)` syscall on Linux, so it's more performant than the plain `cp` command on that platform. cf. Homebrew#17373 (review)

Utils::Cp: Add force_system keyword argument

67f280e

This fixes the test for `UnpackStrategy::Directory`, which needs the `cp` command.

Utils::Cp: Deduplicate SystemCommand invocations

eab1e87

MikeMcQuaid reviewed Jun 10, 2024

View reviewed changes

Library/Homebrew/utils/cp.rb Outdated Show resolved Hide resolved

tesaguri added 2 commits June 11, 2024 20:27

Utils::Cp: Add tests

4edbbfd

Utils::Cp: Rename to Utils::Copy

028cfe1

As per review feedback: Homebrew#17373 (comment)

tesaguri force-pushed the cp-reflink branch from 55d08a4 to 028cfe1 Compare June 12, 2024 03:22

brew style --fix

8e8d0c0

Utils::Copy: Fix tests for generic OS

e4fefc9

MikeMcQuaid reviewed Jun 12, 2024

View reviewed changes

Library/Homebrew/utils/copy.rb Show resolved Hide resolved

extend/os/mac/utils/copy: typed: strict

a30cd15

MikeMcQuaid approved these changes Jun 13, 2024

View reviewed changes

MikeMcQuaid merged commit 1e0add6 into Homebrew:master Jun 13, 2024
25 checks passed

gjpsquare mentioned this pull request Jun 14, 2024

[Build Failure]: cp: invalid option d12frosted/homebrew-emacs-plus#688

Closed

3 tasks

apainintheneck mentioned this pull request Jun 18, 2024

Revert "Use cp -c when copying files" #17526

Merged

khipp mentioned this pull request Jun 22, 2024

temurin@17 and the others are not installing Homebrew/homebrew-cask#177454

Closed

6 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Use `cp -c` when copying files #17373

Use `cp -c` when copying files #17373

tesaguri commented May 27, 2024 •

edited

Loading

Bo98 commented May 27, 2024 •

edited

Loading

tesaguri commented May 27, 2024

Bo98 commented May 27, 2024 •

edited

Loading

tesaguri commented May 27, 2024

Bo98 commented May 27, 2024 •

edited

Loading

MikeMcQuaid commented May 27, 2024

MikeMcQuaid left a comment

tesaguri commented Jun 7, 2024 •

edited

Loading

MikeMcQuaid left a comment

MikeMcQuaid commented Jun 7, 2024

tesaguri commented Jun 8, 2024

Bo98 commented Jun 8, 2024 •

edited

Loading

tesaguri commented Jun 8, 2024 •

edited

Loading

tesaguri commented Jun 9, 2024

Bo98 commented Jun 9, 2024 •

edited

Loading

Bo98 commented Jun 9, 2024

MikeMcQuaid left a comment

tesaguri commented Jun 12, 2024

MikeMcQuaid left a comment

MikeMcQuaid commented Jun 13, 2024 •

edited

Loading

apainintheneck commented Jun 18, 2024

tesaguri commented Jun 18, 2024

apainintheneck commented Jun 18, 2024

MikeMcQuaid commented Jun 19, 2024

jmcbsimmonds commented Jun 22, 2024

apainintheneck commented Jun 22, 2024

MikeMcQuaid commented Jun 24, 2024

Use cp -c when copying files #17373

Use cp -c when copying files #17373

Conversation

tesaguri commented May 27, 2024 • edited Loading

Bo98 commented May 27, 2024 • edited Loading

tesaguri commented May 27, 2024

Bo98 commented May 27, 2024 • edited Loading

tesaguri commented May 27, 2024

Bo98 commented May 27, 2024 • edited Loading

MikeMcQuaid commented May 27, 2024

MikeMcQuaid left a comment

Choose a reason for hiding this comment

tesaguri commented Jun 7, 2024 • edited Loading

Reproduction

MikeMcQuaid left a comment

Choose a reason for hiding this comment

MikeMcQuaid commented Jun 7, 2024

tesaguri commented Jun 8, 2024

Bo98 commented Jun 8, 2024 • edited Loading

tesaguri commented Jun 8, 2024 • edited Loading

tesaguri commented Jun 9, 2024

Bo98 commented Jun 9, 2024 • edited Loading

Bo98 commented Jun 9, 2024

MikeMcQuaid left a comment

Choose a reason for hiding this comment

tesaguri commented Jun 12, 2024

MikeMcQuaid left a comment

Choose a reason for hiding this comment

MikeMcQuaid commented Jun 13, 2024 • edited Loading

apainintheneck commented Jun 18, 2024

tesaguri commented Jun 18, 2024

apainintheneck commented Jun 18, 2024

MikeMcQuaid commented Jun 19, 2024

jmcbsimmonds commented Jun 22, 2024

apainintheneck commented Jun 22, 2024

MikeMcQuaid commented Jun 24, 2024

Use `cp -c` when copying files #17373

Use `cp -c` when copying files #17373

tesaguri commented May 27, 2024 •

edited

Loading

Bo98 commented May 27, 2024 •

edited

Loading

Bo98 commented May 27, 2024 •

edited

Loading

Bo98 commented May 27, 2024 •

edited

Loading

tesaguri commented Jun 7, 2024 •

edited

Loading

Bo98 commented Jun 8, 2024 •

edited

Loading

tesaguri commented Jun 8, 2024 •

edited

Loading

Bo98 commented Jun 9, 2024 •

edited

Loading

MikeMcQuaid commented Jun 13, 2024 •

edited

Loading