Re: Assertion in sort.c

[John Darrington]

OK.  I've checked in patch which partially fixes the problems.

1.  That assertion in merge still needs to allow for max_order ==1
     Otherwise we get a divide by zero condition.

2.  I've added  a signal handler for that event.

3.  I have added a test case test/bugs/big-input-2.sh  which
    demonstrates the outstanding problems.  Unfortunately it takes a
    horribly long time to run.


So the underlying problem remains .... 

I would start by looking at the merge_once function.  There's a lot of
activities going on there which look highly suspicious to me.

J'



On Thu, Jan 06, 2005 at 04:02:17PM +0800, John Darrington wrote:
     On Wed, Jan 05, 2005 at 11:15:44PM -0800, Ben Pfaff wrote:

          Is there a test case to reproduce this?  I can try to do it from
          just a backtrace but a test case makes it easier.
     
     Unfortunately I can only reproduce the problem with one particular
     dataset that has over  100,000 cases.  If I trim it down, the problem
     goes away.
     
     



-- 
PGP Public key ID: 1024D/2DE827B3 
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://wwwkeys.pgp.net or any PGP keyserver for public key.

pgp0b_PSs4Kfs.pgp
Description: PGP signature