[Bug 296987] bin/md5 (md5sum/sha256sum/...): filenames with UTF-8 characters are always octal-escaped regardless of locale (missing setlocale())

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=296987

            Bug ID: 296987
           Summary: bin/md5 (md5sum/sha256sum/...): filenames with UTF-8
                    characters are always octal-escaped regardless of
                    locale (missing setlocale())
           Product: Base System
           Version: 15.1-RELEASE
          Hardware: Any
                OS: Any
            Status: New
          Severity: Affects Many People
          Priority: ---
         Component: bin
          Assignee: [email protected]
          Reporter: [email protected]

## Summary

md5sum, sha256sum, and the other GNU-compatible *sum links always
byte-escape non-ASCII filename bytes (via strnvis(), VIS_CSTYLE |
VIS_OCTAL) even when the process locale is a UTF-8 locale, because
sbin/md5/md5.c never calls setlocale(). strnvis() therefore always
operates in the default "C" locale, treating every UTF-8 continuation
byte as non-printable.

## Reproduction

    $ locale
    LANG=en_US.UTF-8
    LC_CTYPE="en_US.UTF-8"
    ...
    $ echo abc > testÄÖÜ.txt
    $ md5sum testÄÖÜ.txt
    <hash>  test\303\204\303\226\303\234.txt

Expected: filename printed as-is (UTF-8), matching the active locale,
consistent with GNU coreutils' own behavior for a UTF-8 locale.

## Root cause

sbin/md5/md5.c, function safename() (uses strnvis() with VIS_CSTYLE |
VIS_OCTAL) — the file never calls setlocale(). Confirmed via source
inspection of releng/15.1 (commit 0f691888dc56).

## Suggested fix

Add setlocale(LC_CTYPE, ""); (with #include <locale.h>) early in
main(), before any strnvis() call.

## Environment

FreeBSD 15.1-RELEASE-p1, arm64
This did not occur on FreeBSD 14.3, where sbin/md5/md5.c had no
vis()-based filename escaping at all (feature added alongside the
GNU-sum-compatible output modes in 15.x).

-- 
You are receiving this mail because:
You are the assignee for the bug.