2008/9/19, Daniel Stenberg wrote:

> Possibly we could even make it only get used if enabled with a setopt()
> option to prevent users from getting surprises... it seems the current
> problems haven't hit many people as it has been done like this for many
> years, so perhaps (some/many) people can live with the drawbacks?

Probably nobody has been 'surprised' due to the fact that the alarm
function is not checked in the configure script and in consequence
code inside #ifdef HAVE_ALARM is not even compiled.

Some quick archeology in CVS reveals that the 'alarm' check has never
been done in configure. Should the check be introduced?

-- 
-=[Yang]=-