On Tue, 2016-07-26 at 23:53 +0200, Daniel Stenberg wrote: > Thanks a lot, this is super useful and I immediately have an idea why > this was > a bad change. I'm just a bit swamped for the moment but I'll try to > provide a > test patch to fix this within a few days. It was a great exercise learning git bisect, so it was helpful all around.
If it helps, email me the patch and I'll test it and report back to you. FWIW, I did try a quick test of returning CURLE_AGAIN if maxloops was 0 and data was still pending on the connection. This check was done outside the loop controlled by maxloops. The download performance was still greatly degraded. ------------------------------------------------------------------- List admin: https://cool.haxx.se/list/listinfo/curl-library Etiquette: https://curl.haxx.se/mail/etiquette.html
