chore: regenerate THIRD-PARTY-LICENSES.txt - #333
Conversation
| // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | ||
|
|
||
| /// The PUA code points special-cased in the GB18030 encoder. | ||
| pub(crate) static GB18030_2022_OVERRIDE_PUA: [u16; 18] = [ |
There was a problem hiding this comment.
This new encoding_rs block is not a license — it is a chunk of Rust source code from the crate (gb18030_2022.rs: the GB18030_2022_OVERRIDE_PUA / GB18030_2022_OVERRIDE_BYTES tables), with the BSD-3-Clause header appearing as // comments rather than plain text.
This looks like a cargo-about license-file misdetection rather than an intentional attribution. Evidence:
encoding_rs0.8.35 is unchanged from the base commit, yet this is a brand-new stanza — the crate now appears three times in the file (line 611 in the Apache-2.0 group, this block at 2640, and the real BSD-3-Clause text at 2698).- Line 2698 already carries the same BSD-3-Clause text in proper (uncommented) form, so this block adds no attribution coverage.
- Embedding ~40 lines of the dependency's implementation into a legal-attribution artifact is misleading to anyone auditing it, and it will silently churn on every future regeneration.
Suggest excluding this file from cargo-about detection (e.g. a [encoding_rs] clarification / ignore entry in about.toml) and regenerating, so encoding_rs resolves to just the Apache-2.0 and BSD-3-Clause stanzas.
|
|
||
|
|
||
| ------ | ||
| ** wasi; version 0.11.1+wasi-snapshot-preview1 -- https://crates.io/crates/wasi |
There was a problem hiding this comment.
This PR touches only THIRD-PARTY-LICENSES.txt — Cargo.lock, about.toml, and about.hbs are all unchanged. So neither of the two new stanzas is explained by a dependency change:
wasi0.11.1 is still also listed in the plain Apache-2.0 group (line 2120), so it now appears twice under two mutually-exclusive license selections, and the ~225-line Apache-2.0 body is duplicated verbatim just to append the LLVM exception.encoding_rs0.8.35 gained a stanza containing crate source code (see the other comment).
Both point at a change in the generator rather than the dependency graph. Worth noting in the commit message / PR body which cargo-about version produced this output, and confirming the regeneration is reproducible — otherwise this file will churn unpredictably on each run and reviewers have no way to tell an intentional attribution change from a tool regression.
Signed-off-by: David Leong <leongdl@amazon.com>
8cad424 to
8e21a82
Compare
What
Regenerates
THIRD-PARTY-LICENSES.txtwithscripts/check_third_party_licenses.sh --update.Generated file only, 282 added lines, nothing removed.
Why
The
THIRD-PARTY-LICENSES checkjob fails onmainlineas it stands, so it fails on everyopen PR. I hit it on #332, confirmed it is not that branch's doing, and it reproduces on a
clean
mainlinecheckout:No dependency changed.
cargo-aboutdid. The workflow installs it unpinned:which resolves to 0.9.2 today, and 0.9.2 emits every license file a crate ships rather than
one per crate. Two multi-licensed crates gained a second block:
wasi0.11.1+wasi-snapshot-preview1encoding_rs0.8.35Both were already listed at these versions; the content is what grew. Nothing was removed, so
this adds license text we were previously under-reporting.
Root cause worth a follow-up
Pinning
cargo-aboutin the workflow would make this reproducible instead oftime-dependent. As written, any future
cargo-aboutrelease can turn this check red onunrelated PRs, which is what happened here. I left it out of this change to keep the fix
mechanical and quick to merge — happy to send it separately if you want it.
Testing
Run with
cargo-about0.9.2, matching what CI installs today.