Re: [2/2] server: Clear the queue's QS_SMRESULT bit when sending a new message.

[Alexandre Julliard]

Vincent Povirk  writes:
> This keeps the QS_SMRESULT bit in sync with the reply status of the
> most recent sent message, if it happens that the
> previously-most-recent sent message already has a reply.
> wait_message_reply depends on this to work correctly.

I believe you'll also need to set it again when the result is removed.

[2/2] server: Clear the queue's QS_SMRESULT bit when sending a new message.

[Vincent Povirk]

> I believe you'll also need to set it again when the result is removed.

AFAICT the get_message_reply handler assumes that QS_SMRESULT is
already set when it's called, and just resets it when removing a
result if it should be clear. It shouldn't be possible for
get_message_reply to be called when QS_SMRESULT isn't set, because
wait_message_reply always checks for that bit first.

But now that you mention it, get_message_reply does have that 'cancel'
argument, and if for some reason we did cancel a sendmessage before it
has a reply, that bit wouldn't be set when it should be. I'll send a
follow-up.