Get CI working - #63
Merged
Merged
Conversation
This file was added as a placeholder for a future NTT multiplication implementation but has been all-`todo!()` with no callers since it landed. Under `-D warnings` on stable Rust its eight `dead_code` / `unused_variables` errors fail CI. The CHANGELOG entry that flagged the experimental NTT skeleton stays as historical context; the follow-up can re-add this file (or a different scaffold) at that point. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This associated constant on the internal `PrimitiveUnsigned` trait was declared and implemented for every primitive but never read. Stable clippy's `dead_code` lint flags it under the `-D warnings` CI setting. Drop the constant entirely rather than `#[allow(dead_code)]`-suppressing it; the trait is `pub(crate)`, so this is internal cleanup only. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The word-aligned shortcut in `convert::words_to_chunks` computes `end_pos = start_pos + words_per_chunk` without clamping to `words.len()`, so the last chunk panics when the input doesn't fill exactly `words_per_chunk * chunks` source words. Reproduces on `i686-unknown-linux-gnu` (the `Test x86` CI job) for inputs like `0xffffffff000000001`: that value's bit length is 68, so on a 32-bit `Word` target it's 3 source words but `to_chunks(64)` requests `words_per_chunk = 2`. The second chunk's iteration computes `start_pos = 2`, `end_pos = 4`, and `words[2..4]` then panics with `range end index 4 out of range for slice of length 3`. The caller pre-zeroes each chunk's buffer (`Buffer::push_zeros`), so clamping the read and leaving the rest untouched produces the correct value — the trailing zeros get stripped by `Repr::from_buffer`'s `pop_zeros`. The 64-bit and 16-bit jobs stayed green because the same input fits in fewer source words there and the indexing happens to stay in bounds. (Drive-by: rustfmt reflow on the adjacent `buffers.into_iter()...collect()` chain.) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The `test-x86`, `test-no-std`, and `build-aarch64` jobs already pass `--exclude dashu-python`; the `test`, `test-x86_64`, and `clippy` jobs were missing it. pyo3 0.20.3's `pymethods` macro now triggers the `non_local_definitions` lint on stable rustc, which fails `-D warnings` on those three jobs. Add the exclude consistently. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The 1.61/1.64 matrix branches no longer build: dependency updates have moved several transitive dependencies (digest, ryu, serde_json, quote, thiserror, postgres-types, diesel) past those MSRVs. Newer minimal versions cascade unrecoverably without an MSRV-aware resolver, which itself needs cargo 1.84 — too new to commit to as a baseline yet. Bump MSRV to 1.73 (still over two years old) and exercise just one older toolchain in the matrix. The postgres-types/diesel optional features in `dashu-float` still pull deps that need 1.85+ via edition = "2024", so the MSRV branch strips those features at check time. Stable builds continue to compile them via --all-features. The pre-existing `change_version.patch` (which pinned a specific old `diesel_v2`) is no longer needed and is removed. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The test matrix exercises the alternative `Word` widths (16/32/64) by setting `RUSTFLAGS=--cfg force_bits=N`. Rust 1.80 enabled the `unexpected_cfgs` lint by default, so the three `#[cfg(force_bits = "...")]` sites in `integer/src/arch/mod.rs` need an explicit `cargo:rustc-check-cfg=` declaration to keep `-D warnings` happy on stable. The directive is silently ignored by cargo < 1.80, which is fine — the lint it would suppress doesn't exist on those toolchains either. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`_addcarry_u32` / `_addcarry_u64` / `_subborrow_*` were declared `unsafe fn` until Rust 1.81, which made them safe. The `unsafe` block is retained so the MSRV (1.73) build still compiles, and the `unused_unsafe` lint is suppressed so post-1.81 builds don't fail under `-D warnings`. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mechanical cleanup to keep `-D warnings` clean on current stable:
- `mismatched_lifetime_syntaxes`: add explicit `'_` on elided
references in `fmt/`, `memory.rs`, `modular/`, `repr.rs`.
- `needless_lifetimes`: drop unnecessary `'l`/`'r`/`'a` on impl blocks
in `float/src/{add,div,mul}.rs` and `integer/src/div_const.rs`.
- `needless_borrows_for_generic_args`: drop `&` on array literals
passed to `.rfind()` in `float/src/parse.rs`.
- `manual_div_ceil`: switch hand-rolled `(a + b - 1) / b` to
`Int::div_ceil` in float/{fbig,round} and integer mul/root helpers
and `macros/src/parse/common.rs`.
- `missing_transmute_annotations`: turbofish `transmute::<From, To>`
in `convert.rs` and `repr.rs`.
- `doc_overindented_list_items`: reflow doc list indentation in
`repr.rs`.
- `dead_code`: allow on the now-unused `IBig`/`UBig` constants in
`integer/tests/sign.rs`.
- `double_neg` was renamed to the rustc `double_negations` lint;
update the `#[allow(...)]` accordingly in `integer/tests/sign.rs`.
Also includes the `transmute` annotation fix in `integer/src/convert.rs`
(separated from the earlier `words_to_chunks` bug fix).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Owner
|
Thanks a lot for your kind help! I managed to get the MSRV back to 1.68, which is more pleasing to me. A lot of the CI issues are coming from the benchmark crate, which I will exclude from the whole CI later. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This is almost entirely AI generated work, but I've relatively carefully sanity checked it. Please feel free to pick and choose from it as you wish (it's well factored into commits with each individual change), and I will not be even slightly offended if you want to throw it all away.
A lot of it is just fixing lint failures, which required a lot of threading through lifetimes.
Notable decisions that you might reasonably disagree with:
There's also a genuine bugfix in here (words_to_chunks would panic on 32-bit targets).