parse-args.sh is sourced by other scripts (e.g. scripts/packages/build/*.sh). Adding set -e and especially installing/uninstalling INT/TERM/EXIT traps here mutates the caller shell: the final trap - INT TERM EXIT clears any traps set by the parent script, so the parent no longer has its cleanup/kill-on-exit behavior for the rest of execution. Recommend removing trap logic (and ideally set -e) from this shared, sourced helper, or refactor it into a function that returns parsed values without altering global shell options/traps.

This file is sourced by multiple entrypoint scripts. Setting set -e and registering traps in a sourced library changes the caller’s execution environment, and the trap - INT TERM EXIT at the end immediately removes any traps the parent script set before sourcing this file. functions.sh should avoid changing shell options/traps; keep it to function definitions, and let entrypoint scripts own set -e/trap setup.

rust-toolchain.toml pins a stable toolchain, but this script still exports RUSTDOCFLAGS with -Zunstable-options (nightly-only). Removing cargo doc +nightly will make docs generation fail on stable. Either keep using the nightly toolchain for docs, or remove/guard the nightly-only -Z flags so cargo doc works on stable as intended.

exit_code is not initialized before being used in the INT/TERM trap. If the script is interrupted before any command fails, exit $exit_code will typically exit with the status of the last successful command (often 0), masking the interrupt. Initialize exit_code to a non-zero default (e.g. 130) or set explicit exit codes per-signal in the trap handlers.

exit_code is not initialized before being used in the INT/TERM trap. On SIGINT/SIGTERM, this can cause the script to exit 0 (status of the last successful command) instead of a non-zero signal exit code, which can incorrectly mark CI runs as successful. Initialize exit_code up front or use explicit per-signal exit codes in the trap.

The comment says “Remove traps to exit with 0”, but unsetting traps doesn’t control the exit status; it just restores default signal/exit handling. Consider rewording to reflect the actual intent (e.g. avoiding the EXIT trap cleanup on successful completion).

The comment says “Remove traps to exit with 0”, but unsetting traps doesn’t set/force exit status 0; it only removes the trap handlers. Consider rewording to avoid implying exit-code behavior.

The comment says “Remove traps to exit with 0”, but unsetting traps doesn’t influence the exit code; it only restores default signal/exit handling. Consider rewording to match behavior.

Dropping set -e here changes failure semantics: if cargo test --lib/... fails but cargo test --doc passes, this script will exit 0 and mask the failure. Restore set -e (or explicitly check/propagate the exit code after each cargo test) to ensure test failures are not hidden.

The PR title/description says this is only about removing +nightly from the Rust docs build, but this PR also removes set -e / trap-based error handling in multiple unrelated scripts (e.g., build/test scripts). If this broader behavior change is intentional, the PR description should be updated (and ideally split into a separate PR) so reviewers can evaluate the risk and rationale separately.