Partial download crashes with regex parse error

[Whyme Lyu]

Environment

• OS: Arch Linux x86_64
• xh: version 0.20.1. both the Arch official package and the Github release binary xh-v0.20.1-x86_64-unknown-linux-musl.tar.gz can reproduce the issue.

Steps to Reproduce

• Prepare a partially downloaded file (doesn't need to be actual partial file, any non-zero file will do): head -c100 > partial.iso
• Download a file from an HTTP server which supports partial download: xh -p= -dcF --check-status -o partial.iso https://geo.mirror.pkgbuild.com/iso/latest/archlinux-x86_64.iso

Expected Result

Continues download from position 100 bytes. Works like curl -L -C - -f -o partial.iso https://geo.mirror.pkgbuild.com/iso/latest/archlinux-x86_64.iso

Actual Result

xh crashes with the following message:

thread 'main' panicked at src/download.rs:107:5:
called `Result::unwrap()` on an `Err` value: Syntax(
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
regex parse error:
    ^bytes (?P<first_byte_pos>\d+)-(?P<last_byte_pos>\d+)/(?:\*|(?P<complete_length>\d+))$
                              ^^
error: Unicode-aware Perl class not found (make sure the unicode-perl feature is enabled)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
)
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Side note

Actually, this is already fixed with c9162b6. But there's no release available yet.

[Jan Verbeek]

Oh, this one's nasty. We do have a test for this feature, but test builds include the assert_cmd dependency, which includes the predicates dependency, which enables the unicode-perl feature for regex, "fixing" the problem.

Two more things went wrong:

• Using \d instead of [0-9]
• Given that we did use \d, disabling the default features for regex

I wonder if we could use non-test builds for the end-to-end tests. But this feature only has a unit test, so it wouldn't have helped here. And missing features should mostly cause problems at compile time, not runtime, regex is a special case.

#333 was a similar case. We should comb through our list of dependencies to look for other features we do use but don't declare.

[Mohamed Daahir]

This is now fixed in xh v0.21.0