algaos/algaos-ebuild-tree/eclass/cargo.eclass, branch next

cargo.eclass: Separate crate unpacking logic

2026-09-14T17:30:59Z

Split the *.crate file unpacking logic into a separate cargo_crate_unpack() function, so it can be combined with unpacking logic of other eclasses (and ebuilds) without extracting other archives twice (or unnecessarily). Signed-off-by: ingenarel (NeoJesus) Closes: https://bugs.gentoo.org/964342 Patch: https://archives.gentoo.org/gentoo-dev/20260812122836.2302711-1-ingenarel_neojesus@disroot.org/ Signed-off-by: Michał Górny

cargo.eclass: avoid heredoc in global scope

2026-07-10T16:44:50Z

A heredoc may cause the use of a temporary file with < EAPI 9 and even then still with >= EAPI 9 if the input is large. This becomes a problem if sandboxing is applied to metadata generation. Bug: https://bugs.gentoo.org/978857 Signed-off-by: Sam James

cargo.eclass: fix indentation

2026-07-10T16:44:48Z

Signed-off-by: Sam James

cargo.eclass: die if heredoc creation fails

2026-07-08T00:12:25Z

`read ... <<<${x}` will create a heredoc which requires a temporary file depending on the size of the input. Add an explicit `die` if it fails so we get a proper error message under sandbox. Bug: https://bugs.gentoo.org/978857 Signed-off-by: Sam James

cargo.eclass: use direct CDN urls for crates.io

2026-05-29T14:55:07Z

There are four reasons for doing this: - it is the destination of a redirect at the urls we previously used, so less work - per https://blog.rust-lang.org/2024/03/11/crates-io-download-changes/ the API is "slow and expensive" and cargo now uses the CDN directly, too; all third-party "build systems" are recommended to do so as well - crates.io/api has started banning download agents that don't suitably identify themselves, and curl/wget without --user-agent fail. This will also be fixed by https://github.com/gentoo/portage/pull/1585 however that depends on users not manually setting FETCHCOMMAND with naive (no user-agent) values, nor using pk pkg env and copy/pasting. - shorter urls *plus* lack of need for arrow renaming will nontrivially reduce the size of SRC_URI Bug: https://bugs.gentoo.org/682610 Closes: https://bugs.gentoo.org/976165 Reviewed-by: Sam James Signed-off-by: Eli Schwartz

cargo.eclass: add array for skipping tests

2026-04-15T00:25:43Z

Add an array 'CARGO_SKIP_TESTS' that can be declared to skip multiple tests. All test names will be converted to the right command line flag for cargo test. Signed-off-by: idealseal Part-of: https://codeberg.org/gentoo/gentoo/pulls/269 Signed-off-by: Matt Jolly

cargo.eclass: remove ridiculously local einfo

2026-03-31T06:42:57Z

This is something I have locally for verbosity reasons, because I personally dislike warning people to make giant uncached per-CPV tarballs full of other tarballs. Possibly evident from "you are all good"... But my personal taste here should never have been pushed to ::gentoo and I'm at a loss to explain how the diff hunk ended up in a totally unrelated commit. Purge it with fire from the public repo. Sorry... Fixes: 4bf5c8c63455def273c2c27006a1363d23515eb8 Signed-off-by: Eli Schwartz

cargo.eclass: avoid useless use of assert / pipestatus

2026-03-12T15:21:00Z

The preceding command pipeline is `printf | xargs`, and printf is a bash builtin. It cannot fail. `die`ing if xargs fails to consume the pipe buffer is sufficient. assert is banned in EAPI 9, so it needs to go regardless. There's no point to using `inherit eapi9-pipestatus` here. Signed-off-by: Eli Schwartz

cargo.eclass: Document _cargo_set_crate_uris as internal function

2026-02-20T17:23:57Z

Signed-off-by: Ulrich Müller

cargo.eclass: Build with --target for non-native multilib ABIs

2025-12-30T11:26:11Z

Ebuilds were having to explicitly pass --target to work around this. Signed-off-by: James Le Cuirot