IPP backend performance issue

[michaelrsweet]

Version: 1.3-current
CUPS.org User: krumboeck

Between each queued job there exists a 10 second pause when ipp is used. Especially for very fast printers (label printers, etc.) this is a real problem.

In our environment we sometimes queue about 300 jobs at once. The time between the labels shouldn't be larger than 2 seconds.

The problem is caused by the 10-second polling loop in the ipp backend.

[michaelrsweet]

CUPS.org User: mike

Fixed in Subversion repository.

The new code checks at varying intervals of 1 to 10 seconds, starting at 1 second.

[michaelrsweet]

"str2548.patch":

Index: ipp.c

--- ipp.c (revision 7008)
+++ ipp.c (working copy)
@@ -1055,7 +1055,7 @@

 _cupsLangPuts(stderr, _("INFO: Waiting for job to complete...\n"));

• for (; !job_cancelled;)

• for (delay = 1; !job_cancelled;)
  {
  /*

• Build an IPP_GET_JOB_ATTRIBUTES request...
  @@ -1146,10 +1146,14 @@
  check_printer_state(http, uri, resource, argv[2], version, job_id);

  /*

•  \* Wait 10 seconds before polling again...
  

•  * Wait 1-10 seconds before polling again...
  

  */

•  sleep(10);
  

•  sleep(delay);
  

•  delay ++;
  

•  if (delay > 10)
  

•    delay = 1;
  

  }
  }