Re: [Qemu-devel] [PATCH 3/4] doc: add item for "-M enforce-config-section"

[Markus Armbruster]

Eduardo Habkost <address@hidden> writes:

> (CCing Greg, the original author of the code that added the
> enforce-config-section option)
>
> On Tue, Jul 04, 2017 at 10:06:54AM +0200, Markus Armbruster wrote:
>> Peter Xu <address@hidden> writes:
>> 
>> > It's never documented, and now we have one more parameter for it (which
>> > means this one can be obsolete in the future). Document it properly.
>> >
>> > Although now when enforce-config-section is set, it'll override the
>> > other "-global" parameter, that is not necessarily a rule. Forbid that
>> > usage in the document.
>> >
>> > Suggested-by: Eduardo Habkost <address@hidden>
>> > Signed-off-by: Peter Xu <address@hidden>
>> > ---
>> >  qemu-options.hx | 8 ++++++++
>> >  1 file changed, 8 insertions(+)
>> >
>> > diff --git a/qemu-options.hx b/qemu-options.hx
>> > index 297bd8a..927c51f 100644
>> > --- a/qemu-options.hx
>> > +++ b/qemu-options.hx
>> > @@ -85,6 +85,14 @@ Enables or disables NVDIMM support. The default is off.
>> >  @item s390-squash-mcss=on|off
>> >  Enables or disables squashing subchannels into the default css.
>> >  The default is off.
>> > address@hidden enforce-config-section=on|off
>> > +Decides whether we will send the configuration section when doing
>> > +migration. By default, it is turned on. We can set this to off to
>> > +explicitly disable it. Note: this parameter will be obsolete soon,
>> > +please use "-global migration.send-configuration=on|off" instead.
>> 
>> Please say "... is deprecated, please use ...", to make it visible in
>> "git-grep -i deprecat".
>> 
>> > +"enforce-config-section" cannot be used together with "-global
>> > +migration.send-configuration". If it happens, the behavior is
>> > +undefined.
>> 
>> Nasty.  Could we catch and reject such invalid usage?
>
> Actually, the machine option will override
> migration.send-configuration=off, and I don't believe we will break that
> rule.  Documenting that behavior (but warning that it is deprecated)
> sounds easier than adding extra code to detect the conflicting options.
>
> We can simply replace "is undefined"  with "enforce-config-section will
> override migration.send-configuration, but enforce-config-section is
> deprecated".

Better than the scary "behavior is undefined".  But do we have to say
anything at all?  If somebody gives both options with different values,
the conflicting settings fight it out.  In a sane command line, the last
one wins.  Ours isn't sane.  Do we really have to specify who wins?

> I don't think anybody is relying on that option. If nobody is using it,
> we can remove its code soon if we make it trigger a warning.
> Machine-type compatibility code is already using
> migration.send-configuration instead.

What about

      @item enforce-config-section=on|off
      Controls sending of the configuration section when doing migration
      (default on).  Note: this parameter is deprecated, please use
      "-global migration.send-configuration=on|off" instead.