I think I have the answer to your why.

>From multi.c:
    case CURLM_STATE_TOOFAST: /* limit-rate exceeded in either direction */
      /* if both rates are within spec, resume transfer */
      if(Curl_pgrsUpdate(easy->easy_conn)) //!!Progress Callback parent called!!
        easy->result = CURLE_ABORTED_BY_CALLBACK;
      else
        easy->result = Curl_speedcheck(data, now);

      if(( (data->set.max_send_speed == 0) ||
           (data->progress.ulspeed < data->set.max_send_speed )) &&
         ( (data->set.max_recv_speed == 0) ||
           (data->progress.dlspeed < data->set.max_recv_speed)))
        multistate(easy, CURLM_STATE_PERFORM);
      break;

I believe you're calling Curl_pgrsUpdate() here in order to update progress.ulspeed and progress.dlspeed, additionally, you're calling Curl_pgrsUpdate() other places in the code as well, unrelated to speed limiting.

To solve, the parts within Curl_pgrsUpdate() which update progress.ulspeed and progress.dlspeed should probably be factored into a new function. Then this segment in multi.c can call the new function, while other places can continue to call Curl_pgrsUpdate() which in turn calls the new function.

Additionally, I noticed that speed limiting on transfering large files isn't precisely accurate. Meaning I've maxed an upload to 8MB/s, yet external reporting tools are telling me the upload is approaching 9MB/s. This probably means that a moving average somehow isn't being precisely accounted for...

---
** [bugs:#1238] progressfunction called too much**
**Status:** open
**Labels:** progresscallback rate-limiting 
**Created:** Mon Jun 03, 2013 08:55 PM UTC by Daniel Stenberg
**Last Updated:** Tue Jun 11, 2013 06:21 AM UTC
**Owner:** nobody
(bug reported by Miguel Angel in http://curl.haxx.se/mail/lib-2013-05/0254.html)
I am using pycurl in a small script to download coursera videos, it's working 
fine but the callback i installed to display a progress bar in console it's 
called 22188558 times in 70 seconds of execution, it hangs 1 core of the machine 
all the time. 
given the documentation states should be called "roughly once per second or 
sooner" this seems too much for me, i am using this code to use it, it's python 
but i think it's pretty readable.
    ::: python 
    curl = pycurl.Curl() 
    curl.setopt(curl.URL, url) 
    if rate_limit is not None: 
        curl.setopt(curl.MAX_RECV_SPEED_LARGE, rate_limit) 
    file_store = open(filename, "wb") 
    curl.setopt(curl.WRITEDATA, file_store) 
    curl.setopt(curl.NOPROGRESS, 0) 
    curl.setopt(curl.PROGRESSFUNCTION, curl_progress) 
    curl.setopt(curl.FOLLOWLOCATION, 1) 
    #cookies 
    curl.setopt(curl.COOKIEJAR, cookies_filename) 
    curl.setopt(curl.COOKIEFILE, cookies_filename) 
    try: 
        curl.perform() 
    except: 
        import traceback 
        my_logger.error(u"Error downloading file: %s" % traceback.format_exc()) 
    #cleaning 
    curl.close() 
    file_store.close() 
no matter that i define curl_progress as: 
def curl_progress(dl_total, dl_now, ul_total, ul_now): 
    return 0 
it's called so much times that hangs 1 cpu core anyway. 
---
Sent from sourceforge.net because you indicated interest in <https://sourceforge.net/p/curl/bugs/1238/>
To unsubscribe from further messages, please visit <https://sourceforge.net/auth/subscriptions/>