Mercurial > pidgin.yaz
diff libgaim/protocols/oscar/flap_connection.c @ 15092:525607f86cce
[gaim-migrate @ 17877]
Millisecond accuracy for the rolling average thing. It still seems pretty
inaccurate.
committer: Tailor Script <tailor@pidgin.im>
author | Mark Doliner <mark@kingant.net> |
---|---|
date | Sat, 02 Dec 2006 19:57:47 +0000 |
parents | 9cdc8bb39cf2 |
children | b2b0839f57d0 |
line wrap: on
line diff
--- a/libgaim/protocols/oscar/flap_connection.c Sat Dec 02 10:43:06 2006 +0000 +++ b/libgaim/protocols/oscar/flap_connection.c Sat Dec 02 19:57:47 2006 +0000 @@ -89,18 +89,25 @@ if ((snacpair->group == family) && (snacpair->subtype == subtype)) { /* - * We've found the rateclass for this SNAC family and - * subtype! Update our "current" average by calculating - * a rolling average. This is pretty shoddy. We should - * really keep track of the times when the last last - * windowsize messages that were sent and just calculate - * the REAL average. + * We've found the rateclass for this SNAC family + * and subtype! Update our "current" average by + * calculating a rolling average. This is pretty + * shoddy. We should really keep track of the times + * when the last windowsize messages that were sent + * and just calculate the REAL average. */ - time_t now; - now = time(NULL); + struct timeval now; + struct timezone tz; + unsigned long timediff; /* In milliseconds */ + + gettimeofday(&now, &tz); + timediff = MIN((now.tv_sec - rateclass->last.tv_sec) * 1000 + (now.tv_usec - rateclass->last.tv_usec) / 1000, rateclass->max); + /* This formula is taken from the joscar API docs. */ - rateclass->current = MIN(((rateclass->current * (rateclass->windowsize - 1)) + (now - rateclass->last)) / rateclass->windowsize, rateclass->max); - rateclass->last = now; + rateclass->current = MIN(((rateclass->current * (rateclass->windowsize - 1)) + timediff) / rateclass->windowsize, rateclass->max); + rateclass->last.tv_sec = now.tv_sec; + rateclass->last.tv_usec = now.tv_usec; + return; } }