[Bug 44465] Re: I saw the issue on x86. Transiently.

[Richard Theil]

cupsd locked up at roughly 80-90% in top with the remainder going to the drivers. One printer (Epson Stylus C70) had been added on USB before. I was going to configure an HP LaserJet 5L on lp0. gnome-cups-add hangs displaying "reading driver database" for over 3 minutes (on a 1800 MHz Athlon). I did a "/etc/init.d/cupsys stop". As I issued that command, gnome-cups-add continued and showed me both printers were recognized, however, in the next step (selecting the drivers) there was an empty list.
During the hanging, I extracted the following output:

 5734 cupsys    16   0  4344 2120 1440 S 70.6  0.4   1:56.08 cupsd
 5794 cupsys    25   0  9812 7752  716 R 23.5  1.5   0:27.14 cups-driverd
 4526 root      15   0 99268  16m 7964 S  2.0  3.2   0:21.99 Xorg

root at neon:/home/rich# strace -p 5794
Process 5794 attached - interrupt to quit
Process 5794 detached

root at neon:/home/rich# strace -p 5734
...
select(1024, [0 1 2 4 5 6 9], [5 6], NULL, {1, 0}) = 2 (out [5 6], left {1, 0})
time(NULL)                              = 1150204917
time(NULL)                              = 1150204917
time(NULL)                              = 1150204917
...
repeated over and over.

root at neon:/home/rich# lsof | grep 5734
[and then, out of a sudden, gnome-cups-add continued and allowed me to add the printer, with cupsd at 3:24 cpu time. ]
 5734 cupsys    15   0  4344 2120 1440 S  0.3  0.4   3:24.25 cupsd

Was that a coincidence, a heisenbug or did it actually require 800
billion operations to figure out what printers i might have...? On a
retry, it went rather quick (about 10 seconds) and seemed to work. But
IIRC, i have seen the hanging gnome printer discovery briefly on a
completely unrelated PIII-machine with an earlier version of cups (where
I was more concerned about the rip crashing like i reported in another
bug).

-- 
can't add printer, consumes 100% CPU
https://launchpad.net/bugs/44465