Re: [Qemu-block] [Qemu-devel] [PATCH 2/2] Add file-posix-dynamic-auto-read-only feature

[Markus Armbruster]

Peter Krempa <address@hidden> writes:

> On Fri, Mar 29, 2019 at 14:05:46 +0100, Markus Armbruster wrote:
>> Peter Krempa <address@hidden> writes:
>> 
>> [...]
>> > Whether there is a static entry in the QMP schema (which feels wrong
>> > btw as it does not actually expose something which is regarding the
>> > interaction with QMP) or something like this, we need it yesterday. I
>> > don't want to add version checks any more.
>> 
>> I do understand the pressing need for a solution.  But beware of "we
>> need to do something, $this is something, therefore we need to do
>> $this".
>
> Well, then that tends to result in a perpetual search for the optimal
> solution while libvirt is stuck with adding version checks or false
> witnesses (checking that something not entirely relevant, which was
> added at the same time, is present).

"Tends to" yes.  "Does" only if we fail to pick a solution.

All I wanted to say was "beware of first past the post".  I did not mean
to say "all that matters is getting to the post prettily, not when".

> So, is the issue that fixes to code which are always unconditionally
> present would show up both in the new command and in the schema? In the
> case of fixes this will always happen. As I've tried to say in the
> first reply a solution like this feels much better than trying to stuff
> it somehow into the QMP schema without an command. If a fix to the code
> is advertised it does not have much to do with QMP itself, thus adding
> anything to QMP schema only would feel partially odd.

I'm not sure I understand.

The proposed solution puts the information in the schema.  It also makes
it available in the result of the new command.  But why bother running
the new command when you already run query-qmp-schema?