NextJobId incorrect after a cupsd crash

[michaelrsweet]

Version: 1.2-current
CUPS.org User: jlovell

NextJobId is incorrect after a cupsd crash.

In the case where the cache is out of date maybe NextJobId should come from load_request_root() rather than load_next_job_id()...

Steps to reproduce:

• Starting with a clean system:

  killall cupsd

  rm -f /var/spool/cups/{c,d}* /var/tmp/cups/job.cache

  cupsd

• Queue job 1:
  $ lp -H hold main.c
  request id is Lexmark_Z53-1 (1 file(s))

• Cause the job cache to be written:

  killall cupsd

  cupsd

• Queue job 2:
  $ lp -H hold job.c
  request id is Lexmark_Z53-2 (1 file(s))

• SIGKILL cupsd & restart it (with an out-of-date cache):

  killall -9 cupsd

  cupsd

• Queue what should be job 3:
  $ lp -H hold printers.c
  request id is Lexmark_Z53-2 (1 file(s))

• See two job-2 entries:
  $ lpq
  Lexmark_Z53 is not ready
  Rank Owner Job File(s) Total Size
  1st jlovell 1 main.c 57344 bytes
  2nd jlovell 2 job.c 89088 bytes
  3rd jlovell 2 printers.c 88064 bytes

• Confirm the error in the spool dir:

  ls -l /var/spool/cups/

  total 304
  -rw------- 1 root lp 646 Apr 21 14:16 c00001
  -rw------- 1 root lp 654 Apr 21 14:20 c00002
  -rw-r----- 1 root lp 56944 Apr 21 14:16 d00001-001
  -rw-r----- 1 root lp 87196 Apr 21 14:20 d00002-001
  drwxrwx--T 2 root lp 68 Apr 21 13:07 tmp

FWIW this comes from:
rdar://problem/4501801 scheduled jobs are always given job ID "job 1"

Thanks!

[michaelrsweet]

CUPS.org User: mike

How about the dates on the spool directory and cache file?

ls -ld /var/spool/cups
ls -l /var/cache/cups/job.cache

[michaelrsweet]

CUPS.org User: jlovell

ls shows them as equal but gdb reveals the dir is newer:

ls -ld /var/spool/cups /var/tmp/cups/job.cache

drwx--x--- 11 root lp 374 Apr 21 22:07 /var/spool/cups
-rw-r--r-- 1 root lp 229 Apr 21 22:07 /var/tmp/cups/job.cache

(gdb) p dirinfo.st_mtimespec.tv_sec - fileinfo.st_mtimespec.tv_sec
$4 = 12
(gdb) bt
#0 cupsdLoadAllJobs () at job.c:812
#1 0x0002674c in cupsdReadConfiguration () at conf.c:1045
#2 0x000031d8 in main (argc=2, argv=0xbffff88c) at main.c:402

[michaelrsweet]

CUPS.org User: mike

OK, I found the problem - we were using NextJobId from the job.cache file, even if it was less than the highest job ID already seen.

I've added a check in load_next_job_id() for this...