[Libguestfs] [nbdkit] Two POSIX questions for you ...

[Laszlo Ersek]

On 10/9/23 09:46, Richard W.M. Jones wrote:
> Hi Eric, a couple of POSIX questions for you from nbdkit.
> 
> The first question is from an AUR comment on nbdkit:
> 
>   https://aur.archlinux.org/packages/nbdkit#comment-937381
> 
> I think there's a bash-ism in the logscript parameter in this test:
> 
>   https://gitlab.com/nbdkit/nbdkit/-/blame/master/tests/test-log-script-info.sh#L51
> 
> I believe it is happening in the $(( .. )) expression.  How do we
> write that so it'll work in a posix shell?

(I'm not Eric, but curious! :) )

Here I think we should just explicitly insist on bash. Shell arrays are
a bash-specific feature, and the extent array is deeply ingrained. See
especially commit df63b23b6280 ("log: Use strict shell quoting for every
parameter displayed in the log file.", 2021-01-04). In particular the
assignment

extents=(0x0 0x8000 "hole,zero" 0x8000 0x8000 "")

turns "extents" into an array variable.

In the bug tracker, comment
<https://aur.archlinux.org/packages/nbdkit#comment-937375> says, "Thus
this is an upstream issue; their scripts are calling sh when they should
be calling bash". I think that's correct; for logscript=..., we should
require /bin/bash in the manual, and execute the script with /bin/bash
explicitly, not just system().

> 
>  - - -
> 
> Secondly while looking into this I was trying variations on:
> 
>   $ POSIXLY_CORRECT=1 ./nbdkit -fv data '1 2 3' --run 'nbdinfo $uri'
> 
> This doesn't actually cause bash to emulate a posix shell, but it does
> uncover a different bug:
> 
>   nbdkit: error: raw|base64|data parameter must be specified exactly once
> 
> This seems to be happening because getopt_long in wrapper.c behaves
> somehow differently parsing when POSIXLY_CORRECT is set.  However I
> couldn't work out exactly why.
> 
>   https://gitlab.com/nbdkit/nbdkit/-/blob/master/wrapper.c?ref_type=heads#L278
> 
> I guess the wrapper ought to work if POSIXLY_CORRECT is set (?)

IIRC, POSIXLY_CORRECT makes getopt() stop parsing options when the first
non-option argument (i.e., first operand) is reached. So "-f" and "-v"
are taken as options, then the two arguments "data" and '1 2 3' are
taken as operands,  and then "--run" and the rest are taken as operands
as well.

I think the following loop is relevant:

  /* Are there any non-option arguments? */
  if (optind < argc) {
    /* Ensure any further parameters can never be parsed as options by
     * real nbdkit.
     */
    passthru ("--");

    /* The first non-option argument is the plugin name.  If it is a
     * short name then rewrite it.
     */
    if (is_short_name (argv[optind])) {
      const char *language;

      /* Plugins written in scripting languages. */
      if (is_script_plugin (argv[optind], &language)) {
        passthru_format ("%s/plugins/%s/.libs/nbdkit-%s-plugin." SOEXT,
                         builddir, language, language);
        passthru_format ("%s/plugins/%s/nbdkit-%s-plugin",
                         builddir, argv[optind], argv[optind]);
      }
      /* Otherwise normal plugins written in C or other languages that
       * compile to .so files.
       */
      else {
        passthru_format ("%s/plugins/%s/.libs/nbdkit-%s-plugin." SOEXT,
                         builddir, argv[optind], argv[optind]);
      }
      ++optind;
    }

    /* Everything else is passed through without rewriting. */
    while (optind < argc) {
      passthru (argv[optind]);
      ++optind;
    }
  }

With POSIXLY_CORRECT set, the "Everything else is passed through without
rewriting" logic extends to "--run" etc. I'm not sure how that would
lead to the specific error message, though.

Laszlo

> 
> Rich.
>