Re: xargs: doesn't fail on missing terminator (which may have severe consequences)

On 12/9/24 03:34, Christoph Anton Mitterer wrote:
> Hey.
>
> Not sure whether this has been brought up before (at least a quick
> search didn't reveal anything in the bug tracker or the archive).
>
> xargs -0 still executes the command even when no terminating NUL has
> been found in the (last) line.
>
> As shown by e.g. the following script foo.sh:
>     #!/bin/sh
>     printf '/'
>     if [ -n "$1" ]; then
>     	kill -9 "$$"
>     fi
>     printf 'tmp\000'
>
> Which yields:
> $ ./foo.sh | xargs -0 printf '»%s«\n'
> »/tmp«
> $ ./foo.sh kill | xargs -0 printf '»%s«\n'
> »/«

The above test case can be simplified to the cases:

   $ printf 'hello' | xargs -0 printf "<%s>\n"
   <hello>

   $ printf 'hello\0' | xargs -0 printf "<%s>\n"
   <hello>

   $ printf 'hello\0world' | xargs -0 printf "<%s>\n"
   <hello>
   <world>

   $ printf 'hello\0world\0' | xargs -0 printf "<%s>\n"
   <hello>
   <world>

This means xargs(1) does currently not care whether the last entry is
'\0'-terminated or not.

> This, of course, can lead to catastrophic behaviour if the command
> that's piped into xargs is killed at the wrong time.
> Just consider an example like:
> $ ./foo.sh kill | xargs -0 rm -rf
>
>
> I think the proper behaviour would be to do both,
> - not include any non-terminated line (e.g. NUL or LF)
>    though I would say it could execute for all other, properly
>    terminated lines, that have been read up to that point
> - perhaps, if a non-terminated line was found, exit with some special
>    exit status (see below)
>
>
> This problem is even mentioned in the most recent 2024 edition of
> POSIX:
> https://pubs.opengroup.org/onlinepubs/9799919799/utilities/xargs.html
>
>> FUTURE DIRECTIONS
>> A future version of this standard may require that, when the -0
>> option is specified, if the standard input is not empty and does not
>> end with a null byte, xargs ignores the trailing non-null bytes.
>
> I'll ask at the WG, whether it wouldn't be better to give an error
> instead of ignoreing it.

IMO the above, current bahavior is not that terribly bad, but rather
quite good and consistent.

A question in this context may rather be how xargs(1) should handle empty, i.e.,
Zero-length, entries: 'hello\0\0world'.  But I don't see much to improve
or change at this point.

Have a nice day,
Berny