> Well, it would perhaps make sense to also make the progress callback
> get called when libcurl has figured out a function was aborted due to
> a signal. It would also be a very API- and ABI-friendly change.
>
> Wouldn't that fill you need? And to clarify: the progress callback
> can be called before the transfer begins, as it is used to indicate
> progress of the operation and that covers the entire process.

Well I guess so, provided that this behavior is clearly stated and can
be relied upon. The point is to have a function that is called regularly
enough, especially after interrupting the blocking calls, it would
qualify as "progress" too.

By the way, what is the reason for the semantics of polling by slices of
1000 ms, even if nothing happens? Just to make sure and to call the
progress callback? Am I missing something?

-- 
Pierre Ynard
"Une âme dans un corps, c'est comme un dessin sur une feuille de papier."