Re: [Qemu-devel] [PATCH v3 2/7] hmp: Allow help on preconfig commands

[Markus Armbruster]

"Dr. David Alan Gilbert" <address@hidden> writes:

> * Markus Armbruster (address@hidden) wrote:
>> "Dr. David Alan Gilbert (git)" <address@hidden> writes:
>> 
>> > From: "Dr. David Alan Gilbert" <address@hidden>
>> >
>> > Allow the 'help' command in preconfig state but
>> > make it only list the preconfig commands.
>> >
>> > Signed-off-by: Dr. David Alan Gilbert <address@hidden>
>> > Reviewed-by: Peter Xu <address@hidden>
>> > Reviewed-by: Igor Mammedov <address@hidden>
>> > ---
>> >  hmp-commands.hx | 1 +
>> >  monitor.c       | 8 +++++++-
>> >  2 files changed, 8 insertions(+), 1 deletion(-)
>> >
>> > diff --git a/hmp-commands.hx b/hmp-commands.hx
>> > index 0734fea931..8bf590ae4b 100644
>> > --- a/hmp-commands.hx
>> > +++ b/hmp-commands.hx
>> > @@ -15,6 +15,7 @@ ETEXI
>> >          .params     = "[cmd]",
>> >          .help       = "show the help",
>> >          .cmd        = do_help_cmd,
>> > +        .flags      = "p",
>> >      },
>> >  
>> >  STEXI
>> > diff --git a/monitor.c b/monitor.c
>> > index f4a16e6a03..31c8f5dc88 100644
>> > --- a/monitor.c
>> > +++ b/monitor.c
>> > @@ -957,6 +957,10 @@ static void help_cmd_dump_one(Monitor *mon,
>> >  {
>> >      int i;
>> >  
>> > +    if (runstate_check(RUN_STATE_PRECONFIG) && !cmd_can_preconfig(cmd)) {
>> > +        return;
>> > +    }
>> > +
>> >      for (i = 0; i < prefix_args_nb; i++) {
>> >          monitor_printf(mon, "%s ", prefix_args[i]);
>> >      }
>> > @@ -979,7 +983,9 @@ static void help_cmd_dump(Monitor *mon, const 
>> > mon_cmd_t *cmds,
>> >  
>> >      /* Find one entry to dump */
>> >      for (cmd = cmds; cmd->name != NULL; cmd++) {
>> > -        if (compare_cmd(args[arg_index], cmd->name)) {
>> > +        if (compare_cmd(args[arg_index], cmd->name) &&
>> > +            ((!runstate_check(RUN_STATE_PRECONFIG) ||
>> > +                cmd_can_preconfig(cmd)))) {
>> 
>> Why do we need this check in addition to the one in help_cmd_dump_one()?
>
> To gate entire subtables (we've only currently got 'info' and that's enabled,
> anyway, so it never actually triggers).

Something's bothering me around here, but I can't put a finger on it...
let me think aloud.

There's an asymmetry between command execution and help.  The former has
just one check, in monitor_parse_command().  If the command has a
sub-table, and there are arguments, monitor_parse_command() recurses
into the sub-table.  Thus, if the command is unavailable, the
sub-commands are unavailable as well, regardless of their "p" flags.

Help's recursion is different.  It's limited to two levels, unlike
execution.  help_cmd_dump_one()'s check applies to the last level.
help_cmd_dump()'s check applies to the first level.  If there's just one
level, we check twice.  Perhaps less than elegant, but harmless and
simple.

You can't make a sub-command available without also making the command
available.  Makes sense, I guess.  If you try, your "p" flags on the
sub-commands are silently ignored.  That's a bit ugly, but if it doesn't
bother you, I can pretend I didn't see it ;)

> Dave
>
>> >              if (cmd->sub_table) {
>> >                  /* continue with next arg */
>> >                  help_cmd_dump(mon, cmd->sub_table,
> --
> Dr. David Alan Gilbert / address@hidden / Manchester, UK