On Tue, Sep 25, 2007 at 12:52:03PM +0200, Daniel Stenberg wrote:
> Uhm, if that new argument added to the callback is put last in the function
> prototype it should make existing (binary) applications still work fine,
> won't it?
Uh, on x86, perhaps. I really doubt it would hold across all platforms, ABIs
and calling conventions.
> Also, I was thinking that perhaps it would make sense to pass on a handle
> to the callback and then allow a function from within that function to
> extract this info, as then we can easily add more info to become available
> to the callbacks in the future without having to change the API again...
You could probably send a pointer to some struct, which you could then just
make bigger, if you're worried about ABI incompatibility. Not sure how much
it helps you, though.
> but then again, I'm not sure we can come up with more info that'll make
> sense to make it worthwhile go through that kind of slight inconveniance
> for this timeout counter.
TTL information is coming, but I don't think that goes through that
interface.
> Oh, and I did miss the corresponding update of the man page in the patch!
Yes, I'll need to change those at some point. Right now my primary focus is
to fix a really annoying regression that's been creeping up, though -- not
sure if it's in c-ares or in our code. :-)
/* Steinar */
-- Software Engineer, Google NorwayReceived on 2007-09-25