[Pkg-cups-devel] Bug#502100: cups: socket backend hangs in tight select/read loop (larger printouts?)

[Andreas Mohr]

Package: cups
Version: 1.3.9-1                                                            

Hi,

(this did occur with 1.3.8-11 as well)

I have the suspicion that it seems to be larger printouts which make
the socket backend lock up entirely (up to about 10 pages worked fine multiple times,
however trying something larger locked up multiple times).

strace -f -p gives endless:

read(5, ""..., 1024)                    = 0
select(6, [5], [5], NULL, NULL)         = 1 (in [5])
read(5, ""..., 1024)                    = 0
select(6, [5], [5], NULL, NULL)         = 1 (in [5])
read(5,  <unfinished ...>
Process 9753 detached



# ltrace -f -p 9753
--- SIGSTOP (Stopped (signal)) ---
--- SIGSTOP (Stopped (signal)) ---



(gdb) attach 9753
Attaching to program: /usr/lib/cups/backend-available/socket, process 9753
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
0xb7eee5ae in ?? ()
(gdb) bt
#0  0xb7eee5ae in ?? ()
#1  0xb7f6dff4 in ?? ()
#2  0xb7f6c33f in ?? ()
#3  0x00000005 in ?? ()
#4  0xbfe6b768 in ?? ()
#5  0x00000400 in ?? ()
#6  0x00000000 in ?? ()

A "fin" on frame 0 does _never_ return.

disas on any frame address fails (No function contains specified address.).
Corrupt stack??

Remote printer is hplj4000tn:9100, client connected via two routers (one WLAN).
Routers don't display any relevant firewall logs.

_no_ WLAN driver error messages (acx100).

Printout was a PDF file 2-paged via pdftops/psnup/ps2pdf (61 pages).

Admittedly a pretty problematic bug report, if you have some ideas to try,
then just yell.

Thanks,

Andreas Mohr