Use forward slashes in ZIP files; rewrite archive tests; refine error msg by pzhlkj6612 · Pull Request #1534 · svenstaro/miniserve

pzhlkj6612 · 2025-11-28T09:15:29Z

Hello!

I'm running miniserve on Windows. I found that my files in produced ZIP files cannot be correctly extracted on macOS, but Windows Explorer didn't complain about it. After reading a post ¹, I thought I can fix it. Now here's the PR.

I'm new to Rust. Please review my code and help me correct anything wrong.

My test. Run it on Git Bash (untested on WSL):

Bash shell script

#!/bin/bash

set -eu

# positional args
PORT=${1:-8080}

command -v miniserve 2>/dev/null || (echo 'miniserve not found' && exit 1)

playground="$(TMPDIR="$PWD" mktemp -d --suffix="-$(TZ='UTC' date '+%Y%m%dT%H%M%SZ')")"
wwwroot_dir="$playground/wwwroot"
downloaded_dir="$playground/download"
downloaded_file="$downloaded_dir/download.zip"

mkdir -p "$wwwroot_dir" "$downloaded_dir"

for name in 'out' 'out_dir/mid' 'out_dir/mid_dir/in'; do
    mkdir -p "$wwwroot_dir/${name}_dir"
    touch "$wwwroot_dir/${name}_file"
    dd if=/dev/urandom of="$wwwroot_dir/${name}_file" count=10 status=none
done

miniserve "$wwwroot_dir" --port $PORT --interfaces 127.0.0.1 --enable-zip &
pid=$!

curl -s -L "http://127.0.0.1:$PORT/?download=zip" -o "$downloaded_file"

kill $pid

unzip -l "$downloaded_file"

The output before this PR:

  Length      Date    Time    Name
---------  ---------- -----   ----
        0  1980-01-01 00:00   wwwroot\out_dir/
     5120  1980-01-01 00:00   wwwroot\out_file
        0  1980-01-01 00:00   wwwroot\out_dir\mid_dir/
     5120  1980-01-01 00:00   wwwroot\out_dir\mid_file
        0  1980-01-01 00:00   wwwroot\out_dir\mid_dir\in_dir/
     5120  1980-01-01 00:00   wwwroot\out_dir\mid_dir\in_file
---------                     -------
    15360                     6 files

The output for now:

  Length      Date    Time    Name
---------  ---------- -----   ----
        0  1980-01-01 00:00   wwwroot/out_dir/
     5120  1980-01-01 00:00   wwwroot/out_file
        0  1980-01-01 00:00   wwwroot/out_dir/mid_dir/
     5120  1980-01-01 00:00   wwwroot/out_dir/mid_file
        0  1980-01-01 00:00   wwwroot/out_dir/mid_dir/in_dir/
     5120  1980-01-01 00:00   wwwroot/out_dir/mid_dir/in_file
---------                     -------
    15360                     6 files

Before applying this fix, third-party archive utilities, for example PeaZip and 7-Zip, can't understand those backslashes:

Screenshot of the incorrect directory layout in 3rd-party archive utility PeaZip and 7-Zip

However, Windows Explorer always shows the correct directory hierarchy despite this fix:

Screenshot of a multi-layer directory layout in the sidebar of Windows Explorer

.

windows - Zip files expand with backslashes on Linux, no subdirectories - Super User - https://superuser.com/q/1382839 ↩

svenstaro · 2025-11-28T09:19:17Z

Is this actually a problem in the zip crate or how we use it? It doesn't seem correct to me to try to fix it downstream. However, maybe we're holding it wrong. Could you read into the zip crate and see how they deal with this? Surely it's a common problem.

pzhlkj6612 · 2025-11-28T09:42:24Z

Could you read into the zip crate and see how they deal with this?

In progress.

pzhlkj6612 · 2025-11-28T11:58:40Z

Is this actually a problem in the zip crate or how we use it?

This is actually a problem in the zip (namely, zip-rs/zip2) crate.

It doesn't seem correct to me to try to fix it downstream.

I agree. However, waiting for the fix from upstream will be slow, while using my dirty fix will solve it right away.

... Could you read into the zip crate and see how they deal with this? Surely it's a common problem.

Unfortunately, the zip crate write the path to ZIP files as-is for now. And I've tried their "write_dir" example ¹ and found out they write the path in the same way we do.

There are some issues about path handling in that crate:

zip-rs/zip2#242: Zip package created on Windows will generate messy up files after unpack on MacOS (issue closed but who knows the solution)
zip-rs/zip2#272: Path of directories on Windows wouldn't be recognized as directories path on Unix platform

There is also a comment about using 3rd-party library to handle paths instead of std::path in another thread (zip-rs/zip2#376).

https://github.com/zip-rs/zip2/blob/v6.0.0/examples/write_dir.rs#L88 ↩

svenstaro · 2026-01-07T11:17:05Z

Hey, I'd love to get this merged. Sorry for not giving this any attention these last few weeks. Could you please add a test for Windows so we know it won't be broken again in the future?

pzhlkj6612 · 2026-01-07T11:58:10Z

No problem. Let me add necessary tests.

pzhlkj6612 · 2026-01-08T22:35:05Z

Ah-oh, workflows have not been approved. Let's wait for the Windows CI to fail, then I'll commit to revert the latest commit and you can merge it.

pzhlkj6612 · 2026-01-09T07:00:24Z

9c39959

Please, the last two steps.

tests/archive.rs

svenstaro · 2026-01-09T23:15:35Z

It looks like this currently only contains the tests but not the fixes.

Re-apply ad8d3d9 by reverting a4289ff

pzhlkj6612 · 2026-01-10T06:21:44Z

It looks like this currently only contains the tests but not the fixes.

Yep. That was the second to last step - let the test fail on Windows only.

https://github.com/svenstaro/miniserve/actions/runs/20843902950/job/59964978872

I made it. Now we have the fix: 3b59a69 .

src/archive.rs

Cargo.toml

tests/archive.rs

- sort imports - clean dependency - refine comments and English - use long parameter Co-authored-by: Sven-Hendrik Haase <[email protected]>

pzhlkj6612 · 2026-01-10T20:33:01Z

thread 'show_nested_readme_contents::case_7' (2572) panicked at tests\readme.rs:33:5:
assertion failed: parsed_dom.find(Attr("id", "readme-filename")).next().unwrap().text() ==
    filename
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Timing issue? It runs fine before.

svenstaro · 2026-01-10T22:52:58Z

thread 'show_nested_readme_contents::case_7' (2572) panicked at tests\readme.rs:33:5:
assertion failed: parsed_dom.find(Attr("id", "readme-filename")).next().unwrap().text() ==
    filename
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Timing issue? It runs fine before.

Hm yeah, probably a timing issue. :/ You're very welcome to take a look at that if you want.

svenstaro · 2026-01-19T16:01:07Z

Hey, I'd love to get this into the next release of miniserve. Do you think you'll have some time to finish this up soon? Looks like there's only some testing stuff remaining. I can also help you finish this up if you like.

- Universal definitions. - Reuse HTTP client. - Docstring.

pzhlkj6612 · 2026-01-21T07:00:24Z

@svenstaro: sorry for being late. Please see a076fd3 . This is a rewrite.

pzhlkj6612 · 2026-01-23T06:49:34Z

failures:

---- uploading_files_with_invalid_sha_func_is_prevented::case_3_sha128_hash stdout ----
[miniserve stdout] miniserve v0.32.0
[miniserve stdout] Wed, 21 Jan 2026 22:54:46 +0000 [INFO] starting 3 workers
[miniserve stdout] Bound to [::]:50404, 0.0.0.0:50404
[miniserve stdout] Serving path /private/var/folders/kg/7q73ww8s3llgyl61c9z_j5g40000gn/T/.tmpS6jASu
[miniserve stdout] Available at (non-exhaustive list):
[miniserve stdout]     http://127.0.0.1:50404/
[miniserve stdout]     http://192.168.64.23:50404/
[miniserve stdout]     http://[::1]:50404
[miniserve stdout] 
[miniserve stdout] Wed, 21 Jan 2026 22:54:46 +0000 [INFO] Actix runtime found; starting in Actix runtime
[miniserve stdout] Wed, 21 Jan 2026 22:54:46 +0000 [INFO] starting service: "actix-web-service-[::]:50404", workers: 3, listening on: [::]:50404
[miniserve stdout] Wed, 21 Jan 2026 22:54:46 +0000 [INFO] starting service: "actix-web-service-0.0.0.0:50404", workers: 3, listening on: 0.0.0.0:50404
[miniserve stdout] Wed, 21 Jan 2026 22:54:46 +0000 [INFO] ::1 "GET / HTTP/1.1" 200 33177 "-" "-" 0.001141
Error: rve stderr] Wed, 21 Jan 2026 22:54:46 +0000 [ERROR] Invalid HTTP request
Error: rve stderr] Wed, 21 Jan 2026 22:54:46 +0000 [ERROR] caused by: Invalid header value found for 'X-File-Hash-Function'. Supported values are SHA256 or SHA512. Found SHA128.
[miniserve stdout] Wed, 21 Jan 2026 22:54:46 +0000 [INFO] ::1 "POST /upload?path=/ HTTP/1.1" 400 139 "-" "-" 0.000178
Error: reqwest::Error { kind: Body, url: "http://localhost:50404/upload?path=/", source: SendError { kind: Disconnected } }

Bruh, it works on my machine ™.

pzhlkj6612 · 2026-01-30T14:24:03Z

Hi, @svenstaro ! Does the current revision look good?

svenstaro · 2026-02-08T00:56:56Z

src/archive.rs

+
+            // Workaround for Windows path in ZIP files:
+            // Always use forward slashes for all paths to avoid literal backslashes in saved entry path.
+            // TODO: remove this when the "zip" cargo has the ability to process a directory as a whole.


According to zip-rs/zip2#376 and the release notes of zip v7.2.0, this is now handled by upstream. At least I see they now have tests handling this. Could you check out whether that now does the job? It's a bit odd though that they have that issue still open.

a06e31e. It seems to work. I'll test it soon.

a06e31e. It seems to work. I'll test it soon.

It still does not work on macOS (#1534 (comment)). Reverted by 9f8b0de and a new comment has been added.

svenstaro

Tiny nit but apart from that I really like these tests. I was at first a bit skeptical about the ArchiveKind thing as it adds some complexity but I think here it actually makes sense in the context of the archives_links_and_downloads test.

tests/archive.rs

svenstaro · 2026-02-08T04:10:51Z

Happy to merge after you figure out whether upstream's fix works for us. :)

svenstaro · 2026-02-10T05:55:30Z

This is the last PR before releasing v0.33.0 :)

The 2nd parameter of "assert!()" has "format!()" integrated. Co-authored-by: Sven-Hendrik Haase <[email protected]>

Our dependency "zip" has fixed a bug about setting local platform in the ZIP since v7.2.0: zip-rs/zip2@9e9badd, so we don't need to fix it here. This effectively reverts the main part of commits: - The fix: ad8d3d9 (the origin) or 3b59a69 (the re-applied one) - The test: 90c5234

svenstaro · 2026-02-11T12:23:39Z

Is this still draft or shall I review it?

pzhlkj6612 · 2026-02-11T13:47:02Z

Sorry for the delay. I didn't have a macOS with me several hours ago.

Unfortunately, on macOS, Finder does not treat those backslashes as path separators even though we have already correctly set the field after switching to the newest "zip" package:

Screenshot of wrong structure of a ZIP file on macOS

$ zipinfo -v .../wwwroot.zip | grep 'file system'
  file system or operating system of origin:      MS-DOS, OS/2 or NT FAT
  minimum file system compatibility required:     MS-DOS, OS/2 or NT FAT
  ...

BEFORE:

$ zipinfo -v .../wwwroot.zip | grep 'file system'
  file system or operating system of origin:      Unix
  minimum file system compatibility required:     MS-DOS, OS/2 or NT FAT
  ...

However, both PeaZip and 7-Zip agree with us:

Screenshot of correct structure of a ZIP file in PeaZip and 7-Zip on Windows

Path parsing in ZIP looks implementation dependent. Since we have this:

File name: the name of the file including an optional relative path. All slashes in the path should be forward slashes '/'.

from: The structure of a PKZip file - https://users.cs.jmu.edu/buchhofp/forensics/formats/pkzip.html

I think using forward slashes is necessary and more safe. Let me re-revert the code. Sorry for the noise.

Btw, the "zip" package works as expected: it names the saved item exactly what the user specified.

svenstaro · 2026-02-11T13:49:41Z

Some really peculiar behavior from zip here. Thanks for looking into this!

macOS: rememeber, no forward slashes. This reverts commit a06e31e.

src/archive.rs

svenstaro · 2026-02-15T08:38:57Z

tests/archive.rs

+fn fetch_index_document(
+    reqwest_client: &Client,
+    server: &TestServer,
+    expected: StatusCode,
+) -> Result<Document, Error> {
+    let resp = reqwest_client.get(server.url()).send()?;
+    assert_eq!(resp.status(), expected);
+
+    Ok(Document::from_read(resp)?)
+}


I think it would probably make sense to make this a universal utility and then use it everywhere? We have sooo many Document::from_reads in all the tests that this might be a great follow-up to tidy up some test code if you're up for it. At any rate, this should probably go into tests/utils/mod.rs for now.

Actually, let's do that in a follow-up. I wanna get this one merged at long last.

oops, I'm coding for this. Thank you for reviewing my PR. I think this cleanup will be made in following PRs after the upcoming new verison.

oops Co-authored-by: Sven-Hendrik Haase <[email protected]>

Fix Windows path in ZIP files by using forward slashes only

ad8d3d9

pzhlkj6612 marked this pull request as draft November 28, 2025 09:28

pzhlkj6612 marked this pull request as ready for review November 28, 2025 11:58

pzhlkj6612 marked this pull request as draft January 7, 2026 11:32

pzhlkj6612 added 3 commits January 8, 2026 22:25

Add path of failed files to error message; update comments

160acdc

Test archives with broken symlinks; test Windows path in ZIP files

90c5234

Revert changes to let test fail

a4289ff

Fix failed TAR test - directory walking result varies by OS

9c39959

svenstaro requested changes Jan 9, 2026

View reviewed changes

tests/archive.rs Outdated Show resolved Hide resolved

tests/archive.rs Outdated Show resolved Hide resolved

pzhlkj6612 added 3 commits January 10, 2026 06:04

Merge "master"; use reqwest_client instead

49c05d8

Shrink imports; fix English typo; refine "test_tar_archives"

2fc9059

Fix Windows path in ZIP files by using forward slashes only

3b59a69

Re-apply ad8d3d9 by reverting a4289ff

svenstaro requested changes Jan 10, 2026

View reviewed changes

Apply review suggestions

dc9fc7e

- sort imports - clean dependency - refine comments and English - use long parameter Co-authored-by: Sven-Hendrik Haase <[email protected]>

Rewrite archive tests with throughout coverage

a076fd3

- Universal definitions. - Reuse HTTP client. - Docstring.

pzhlkj6612 marked this pull request as ready for review January 21, 2026 12:35

svenstaro reviewed Feb 8, 2026

View reviewed changes

pzhlkj6612 marked this pull request as draft February 8, 2026 03:00

svenstaro approved these changes Feb 8, 2026

View reviewed changes

tests/archive.rs Outdated Show resolved Hide resolved

tests/archive.rs Outdated Show resolved Hide resolved

svenstaro mentioned this pull request Feb 10, 2026

Archive creation fails if there is a broken symlink #1441

Open

pzhlkj6612 and others added 3 commits February 11, 2026 08:23

Merge "master"

918e2b3

Apply code review suggestions

f38cfde

The 2nd parameter of "assert!()" has "format!()" integrated. Co-authored-by: Sven-Hendrik Haase <[email protected]>

pzhlkj6612 changed the title ~~Fix Windows path in ZIP files by using forward slashes only~~ Rewrite all tests for archives; add filename to error msg from archiving Feb 11, 2026

Reapply "using forward slashes in ZIP on Windows" and a related test

9f8b0de

macOS: rememeber, no forward slashes. This reverts commit a06e31e.

pzhlkj6612 marked this pull request as ready for review February 11, 2026 14:02

pzhlkj6612 changed the title ~~Rewrite all tests for archives; add filename to error msg from archiving~~ Use forward slashes in ZIP files; rewrite archive tests; refine error msg Feb 11, 2026

svenstaro reviewed Feb 15, 2026

View reviewed changes

src/archive.rs Show resolved Hide resolved

svenstaro requested changes Feb 15, 2026

View reviewed changes

Fix wrong comments about slashes

1fe6167

oops Co-authored-by: Sven-Hendrik Haase <[email protected]>

svenstaro merged commit e068024 into svenstaro:master Feb 16, 2026
18 checks passed

svenstaro added a commit that referenced this pull request Feb 16, 2026

Add CHANGELOG entry for #1534

911aef1

Uh oh!

Conversation

pzhlkj6612 commented Nov 28, 2025

Footnotes

Uh oh!

svenstaro commented Nov 28, 2025

Uh oh!

pzhlkj6612 commented Nov 28, 2025

Uh oh!

pzhlkj6612 commented Nov 28, 2025

Footnotes

Uh oh!

svenstaro commented Jan 7, 2026

Uh oh!

pzhlkj6612 commented Jan 7, 2026

Uh oh!

pzhlkj6612 commented Jan 8, 2026

Uh oh!

pzhlkj6612 commented Jan 9, 2026

Uh oh!

Uh oh!

Uh oh!

svenstaro commented Jan 9, 2026

Uh oh!

pzhlkj6612 commented Jan 10, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

pzhlkj6612 commented Jan 10, 2026

Uh oh!

svenstaro commented Jan 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

svenstaro commented Jan 19, 2026

Uh oh!

pzhlkj6612 commented Jan 21, 2026

Uh oh!

pzhlkj6612 commented Jan 23, 2026

Uh oh!

pzhlkj6612 commented Jan 30, 2026

Uh oh!

svenstaro Feb 8, 2026

Choose a reason for hiding this comment

Uh oh!

pzhlkj6612 Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

pzhlkj6612 Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

svenstaro left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

svenstaro commented Feb 8, 2026

Uh oh!

svenstaro commented Feb 10, 2026

Uh oh!

svenstaro commented Feb 11, 2026

Uh oh!

pzhlkj6612 commented Feb 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

svenstaro commented Feb 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

svenstaro Feb 15, 2026

Choose a reason for hiding this comment

Uh oh!

svenstaro Feb 16, 2026

Choose a reason for hiding this comment

Uh oh!

svenstaro commented Jan 10, 2026 •

edited

Loading

pzhlkj6612 commented Feb 11, 2026 •

edited

Loading

svenstaro commented Feb 11, 2026 •

edited

Loading