windows-targets v0.52.3 increased MSRV to 1.60

[lopopolo]

Summary

windows-targets v0.52.3 increased its MSRV in a patch release from Rust 1.56 to Rust 1.60. This means that windows-sys v0.52.0 no longer builds on Rust 1.56, which is its stated MSRV on crates.io and in Cargo.toml.

Crate manifest

[package]
name = "known-folders"
version = "1.1.0" # remember to set `html_root_url` in `src/lib.rs`.
authors = ["Ryan Lopopolo <[email protected]>"]
license = "Apache-2.0 OR MIT"
edition = "2021"
rust-version = "1.58.0"
readme = "README.md"
repository = "https://github.com/artichoke/known-folders-rs"
documentation = "https://docs.rs/known-folders"
homepage = "https://github.com/artichoke/known-folders-rs"
description = "A safe wrapper around the Known Folders API on Windows"
keywords = ["app_dirs", "known-folder", "path", "windows"]
categories = ["api-bindings", "filesystem", "os::windows-apis"]
include = ["examples/**/*", "src/**/*", "tests/**/*", "LICENSE-*", "README.md"]

[dependencies]

[target.'cfg(windows)'.dependencies.windows-sys]
version = "0.52.0"
features = [
  "Win32_Foundation",
  "Win32_Globalization",
  "Win32_System_Com",
  "Win32_UI_Shell",
]

# Check that crate versions are properly updated in documentation and code when
# bumping the version.
[dev-dependencies.version-sync]
version = "0.9.4"
default-features = false
features = ["markdown_deps_updated", "html_root_url_updated"]

[package.metadata.docs.rs]
rustdoc-args = ["--cfg", "docsrs"]
targets = [
  # Tier 1
  "x86_64-pc-windows-msvc",
  "x86_64-pc-windows-gnu",
  "i686-pc-windows-msvc",
  "i686-pc-windows-gnu",
  # Tier 2
  #  "aarch64-pc-windows-msvc",
  #  "i586-pc-windows-msvc",
]

Crate code

No response

[lopopolo]

These are the build logs from my weekly CI build with Rust 1.58 for known-folders:

$ rustc -Vv
rustc 1.58.0 (02072b482 2022-01-11)
binary: rustc
commit-hash: 02072b482a8b5357f7fb5e5637444ae30e423c40
commit-date: 2022-01-11
host: x86_64-pc-windows-msvc
release: 1.58.0
LLVM version: 13.0.0
$ cargo version --verbose
cargo 1.58.0 (7f08ace4f 2021-11-24)
release: 1.58.0
commit-hash: 7f08ace4f1305de7f3b1b0e2f76]5911957226bd4
commit-date: 2021-11-24
host: x86_64-pc-windows-msvc
libgit2: 1.3.0 (sys:0.13.23 vendored)
libcurl: 7.80.0-DEV (sys:0.4.51+curl-7.80.0 vendored ssl:Schannel)
os: Windows 10.0.20348 (Windows Server 2022 Datacenter) [64-bit]
$ cargo build --verbose
    Updating crates.io index
 Downloading crates ...
  Downloaded windows-targets v0.52.3
  Downloaded windows_x86_64_msvc v0.52.3
  Downloaded windows-sys v0.52.0
error: package `windows_x86_64_msvc v0.52.3` cannot be built because it requires rustc 1.60 or newer, while the currently active rustc version is 1.58.0
Error: Process completed with exit code 1.

[kennykerr]

Took a moment to recall, but this is another case where a Cargo workspace bug (rust-lang/cargo#10623) forced an artificial version bump (#2815) as a workaround.

[kennykerr]

I could remove the "rust-version" from the Cargo.toml file and republish. I can't build within the repo with Rust 1.56 because of the Cargo bug, but perhaps it will work as a package. I just don't know how to test that. @ChrisDenton may know.

[kennykerr]

Alternatively, I may just get rid of the workspace as it has caused so many of these kinds of issues.

[ChrisDenton]

You can do a dry run publish with the latest toolchain, e.g.:

cargo publish --dry-run -p windows_x86_64_msvc
cargo publish --dry-run -p windows-targets
git switch --detach 0.52.0
cargo publish --dry-run -p windows-sys

Etc, etc. Publishes to target/package. Then it can be tested with a separate crate using path dependencies and whatever rustc version you like. Admittedly this is a bit of a pain but either removing or restoring the rust-version to 1.56 seemed to work for me, after editing to use path dependencies.

[kennykerr]

Thanks Chris - I think I can just remove the rust-version for now to fix this issue - but after that I may just rip out the workspace.

[kennykerr]

I've published another update and everything seems to be working well:

D:\git\scratch>rustup default
1.56-x86_64-pc-windows-msvc (default)

D:\git\scratch>cargo check       
   Compiling windows_x86_64_msvc v0.52.4
    Checking windows-targets v0.52.4
    Checking windows-sys v0.52.0
    Checking scratch2 v0.1.0 (D:\git\scratch)
    Finished dev [unoptimized + debuginfo] target(s) in 0.46s

[ijackson]

Thanks for fixing the accidental MSRV increase.

However, I don't think yanking 0.52.3 was helpful. For us, it resulted in a hiccup in the release process of Arti. Generally it's not a good idea to yank a crate unless the consequences of a user using the questionable version is worse than a compile error. (Assuming you have published an fix as a point release.)

After all, in this situation, yanking causes a compile error if it does anything at all. And it can create compile errors where in fact everything would have been fine. So yanking a crate to prevent compile errors doesn't make sense.

[kennykerr]

Thanks for the feedback. Yes, that makes sense. I assume at this point there's no value in unyanking? 🤷

[ijackson]

Thanks for the feedback. Yes, that makes sense. I assume at this point there's no value in unyanking? 🤷

Well, I don't think it would do any harm and if someone hasn't noticed a failure yet it might mean they never see it. So I think I would suggest unyanking.

Thanks :-)

[kennykerr]

Done!