Re: [RFC/INCOMPLETE PATCH 0/8] Attempt to make qemu-img options consistent and --help working

[Michael Tokarev]

07.02.2024 22:01, Manos Pitsidianakis:
> Hello Michael,
>
> Such changes are long overdue. However given the complexity of
> commands and arguments, maybe it'd be a good idea to write a code
> generator for the command line interface, This way you could also
> generate --help outputs, manpages, shell completions just from the
> command line spec we'd use to generate the argv parsing routines.

I tried starting with just a set of common options in --help output,
but quickly abandoned that idea.  Even there, and I mentioned that
in the email you're replying to, we should have slightly different
text in different contexts.  So it seems like it's better to just
write it in two places.  Two *different* places, - which is the
--help output and qemu-img.rst documentation (from which the manpage
is generated).  The two places are really different, because --help
is just a very brief usage/options summary, while the doc is a much
more complete descriptive guide.

There's one more context, - the option parsing code. There are ways
to make it easier, like libpopt for example, but in my view, in an
attempt to make life easier, it makes it even more complex.  In my
taste anyway, YMMV.

In short, - while this stuff seems like a good candidate for some
automation, it might be not, and the first step IMHO is to get the
first task done, - to review and refresh all options, see what can
be done with the messy difference of the meanings for subcommands,
describe them.  Maybe after that whole thing can be automated (if
it's possible to do with this differently-named hell and with
readable output).

/mjt