Re: [Qemu-devel] [PATCH 06/10] qapi: simplify qmp_input_next_list

[Paolo Bonzini]

Il 22/03/2012 21:22, Anthony Liguori ha scritto:
>>
>> Signed-off-by: Paolo Bonzini<address@hidden>
> 
> I've been staring a this patch for the past 5 minutes and I can't figure
> out what's going on here.
> 
> Maybe the code was too obscure to begin with.  Could you enhance the
> commit message a bit with what's going on here?

There's three possible questions about what's going on:

1) What's going on before the patch with so->entry

2) What's going on after the patch with so->entry

3) What's going on with *list.  The patch doesn't change this, it just
unties it with so->entry, but it's the most puzzling part so the
question is a good one. :)


First of all, so->entry here is advanced and also used to figure out
whether we have a next element.  It is then accessed in qmp_input_get_obj.

The caller must:

* call start_list

* call next_list for each element *including the first*

* on the first call to next_list, the result is the head of the list
(works for both input and output visitor).

Before: so->entry is initialized to qlist_first on start_list.  The
first call will have *list == NULL, so it skips the qlist_next.  The
caller assigns the result and makes *list not NULL, so that the next
iteration will advance so->entry and modify (*list)->next.

After: so->entry is initialized to NULL on start_list.  The first call
sees so->entry == NULL and does qlist_first; subsequent calls do
qlist_next.  *list is handled same as above: assignment done by the
caller on the first call, done by next_list on the next ones.


Thanks for making me write all this down, because it looks like there is
still room for further simplifying the handling of *list (for example,
do not have the caller write the list head).  I'll submit a v2.
Depending on the amount of simplification I'll document the result in
either a code comment or the commit message.

Luckily this code is very well tested.  It broke all the time for me. :P

Paolo