[Cbc] ./cbc -thread 4 -slog 63 -miplib Segfaults

[John J Forrest]

Sven,

I will get round to looking at it.  I did to a lot of testing on threaded 
version but maybe some recent changes have semi-broken it.

May take some time.  I probably should work out some logging mode so that 
the run can be replayed with solves finishing in same order, but even that 
may not find it.

John



Sven de Vries <devries at ma.tum.de> 
Sent by: cbc-bounces at list.coin-or.org
07/27/2007 04:05 PM

To
Cbc <cbc at list.coin-or.org>
cc

Subject
[Cbc] ./cbc -thread 4 -slog 63 -miplib Segfaults






Hi,

with an updated trunk-version of cbc and threads enabled, I encounter on 
sun a segfault:



% ./cbc -thread 4 -slog 63 -miplib > t44c.log
   processing mps file: 10teams (1 out of 44)
cbc_clp 924 = 924 ; okay - took 325.85 seconds.
   processing mps file: air03 (2 out of 44)
cbc_clp 340160 = 340160 ; okay - took 40.36 seconds.
   processing mps file: bell3a (3 out of 44)
cbc_clp 878430 = 878430 ; okay - took 49.8826 seconds.
   processing mps file: bell5 (4 out of 44)
Segmentation fault


the last lines of that logfile are:

Cbc0038I Mini branch and bound improved solution from 9.07211e+06 to 
9.06465e+06
Cbc0038I Freeing continuous variables gives a solution of 9.06465e+06
Cbc0038I Round again with cutoff of 9.06465e+06
Cbc0038I Pass   3: obj.    0.44490 --> up =     0 , down =     0 -- rand = 
   9 (  25)
Cbc0038I Pass   4: obj.   -0.00000 --> up =     0 , down =     0 -- rand = 
   0 (  20)  - solution found of 8.94337e+06 - not good enough after mini 
branch and bound
Cbc0038I No solution found this major pass
Cbc0038I Before mini branch and bound, 26 integers at bound fixed and 14 
continuous of which 1 were internal integer and 0 internal continuous
Cbc0038I Mini branch and bound improved solution from 8.94337e+06 to 
9.04646e+06
Cbc0012I Integer solution of 9.04646e+06 found by feasibility pump after 0 
iterations and 0 nodes (1.48 seconds)


the crash happens on bell5. it seems not to be fully reproducible and i 
haven't managed (yet) to crash it with debug-binaries ...

also, a hundred calls to
./cbc ../../Data/miplib3/bell5.gz -thread 4 -slog 63 -solve
didn't trigger it.

another attempt with
foreach i ( 0 1 2 3 4 5 6 7 8 9  )
foreach j ( 0 1 2 3 4 5 6 7 8 9  )
echo $i $j
time ./cbc ../../Data/miplib3/bell5.gz -thread 4 -solve > b${i}${j}.log
end
end

however was succesfull:
..
1 1
69.92u 22.99s 0:31.46 295.3%
1 2
57.59u 18.76s 0:26.21 291.3%
1 3
75.86u 24.46s 0:33.58 298.7%
1 4
113.93u 38.68s 0:49.42 308.8%
1 5
Segmentation fault
84.84u 30.05s 0:37.83 303.7%
1 6
68.14u 22.75s 0:30.32 299.7%
1 7

but

grep Result b1[3456].log
b13.log:Result - Finished objective 8.96641e+06 after 9821 nodes and 20296 
iterations - took 75.7604 seconds
b14.log:Result - Finished objective 8.96641e+06 after 18564 nodes and 
33605 iterations - took 113.851 seconds
b15.log:Result - Finished objective 8.96641e+06 after 12041 nodes and 
23725 iterations - took 84.7418 seconds
b16.log:Result - Finished objective 8.96641e+06 after 7435 nodes and 16379 
iterations - took 68.0419 seconds

yields no anomalies. so in this case it seems to be that cbc segfaults 
only while exiting...

best
sven

ps: here is my environment:
uname -m = sun4u 
uname -r = 5.8 
uname -s = SunOS 
uname -v = Generic_117350-31
/usr/bin/uname -p = sparc 
/bin/uname -X     = System = SunOS 
Release = 5.8 
KernelID = Generic_117350-31 
Machine = sun4u 
BusType = <unknown> 
Serial = <unknown> 
Users = <unknown> 
OEM# = 0 
Origin# = 1

...
configure:2949: C compiler options are: -xO4  -DNDEBUG
configure:3078: checking for CC
configure:3094: found /opt/SUNWspro/bin/CC
cc: Sun C 5.8 2005/10/13
CC: Sun C++ 5.8 2005/10/13
_______________________________________________
Cbc mailing list
Cbc at list.coin-or.org
http://list.coin-or.org/mailman/listinfo/cbc

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/cbc/attachments/20070728/e8c73cd3/attachment.html