Added a note on failed handles not being counted by curl_multi_perform.

[akashihi]

No description provided.

[bagder]

Thanks for your help in improving the docs!

[bagder]

1. "a transfer" or "transfers" (not "transfer")
2. Why is this likely to happen "especially" in the case you reuse handles? In my view, transfers are always roughly equally likely to fail.

[akashihi]

1. Fixed
2. Well, you are right, fixed.

[bagder]

This is incorrect. The number of "running handles" is increased by curl_multi_add_handle so a slow start doesn't result in a lower number. The number is however decreased when the transfer is complete, which can happen at any point and thus already in the fist call to curl_multi_perform().

[akashihi]

Rephrased to be more precise.

[bagder]

How about instead trying to not duplicate what the curl_multi_perform.3 man page already says ? We can instead rather remove the (possibly slightly misleading) sentence "Tracking when this number changes, you know when one or more transfers are done."

[akashihi]

Agree

[bagder]

Thanks!