scripts/release.sh now stops before cargo release in two cases where git-cliff gives a misleading answer. In a shallow clone, git-cliff may not see the newest tag, and then it counts from an older tag or from the start and prints a plausible but wrong version. The script says to run git fetch --unshallow --tags first, as didbot's does. And with nothing since the last tag, git-cliff names that tag's version again, so the script now says there is nothing to release instead of running cargo release against an existing tag.
I checked both against copies of this repository: a --depth 1 clone is refused, and a full clone at v0.4.0 with no new commits says nothing to release; v0.4.0 is already tagged. Both exit before cargo release runs.
🤖 Generated with Claude Code