Re: [Qemu-devel] [PATCH] usb-mtp: fix return status of delete

[Bandan Das <bsd@redhat.com>]

Peter Maydell <peter.maydell@linaro.org> writes:

> On Mon, 11 Mar 2019 at 16:14, Bandan Das <bsd@redhat.com> wrote:
>>
>> Peter Maydell <peter.maydell@linaro.org> writes:
>> > At the moment PARTIAL_DELETE is "ALL_DELETE | READ_ONLY", which
>> > doesn't seem like it makes much sense.
>> >
>>
>> Sorry, can you please clarify what doesn't make sense ?
>
> Generally, if you have multiple bits X, Y in a return
> value, they should be independent. Sometimes we define
> a convenience value Z that's X | Y, but then Z should
> have a name that indicates that it's really doing both
> X and Y (for instance often a READWRITE constant will
> be READ | WRITE). In this case, I don't see why
> PARTIAL_DELETE would be a sensible name to indicate
> "both ALL_DELETE and also READ_ONLY" -- if we only
> partially did a delete why do we set the ALL_DELETE bit ?
>

Because during a recursive call, we were able to successfully
delete objects(s) for the previous call but for "this"
set of objects, it failed which is supposed to return a
partial_delete back.

Does simply "DELETE" instead of "ALL_DELETE" seem less
confusing ? I definitely want to keep PARTIAL_DELETE the
way it is simply because it's easier to refer back
to the spec that way.

> It might be useful to take a step back -- what are
> the different possible outcomes from this function that
> we need to distinguish, and when should we be returning
> which outcome?
>
> thanks
> -- PMM