INN 2.4.0 dumps core on Solaris 9

Nathan Coraor nate at cse.psu.edu
Tue Oct 21 20:00:21 UTC 2003


Hello,

I've scoured the web and banged my head against the wall for days
regarding this one and can't come up with a solution.  My problem seems
to be this:

  http://www.isc.org/ml-archives/inn-workers/2000/02/msg00098.html
  http://groups.google.com/groups?selm=8rcqqa%241qv%241%40ultranews.duc.auburn.edu&oe=utf-8&output=gplain

Basically the story is that everything compiles okay, however, innd
dumps core on startup.  A truss reveals:

  29622:  open64("/var/news/etc/newsfeeds", O_RDONLY)     = 18
  29622:  fstat64(18, 0xFFBFFA70)                         = 0
  29622:  brk(0x002F7E00)                                 = 0
  29622:  brk(0x002F9E00)                                 = 0
  29622:  read(18, " # #     $ I d :   n e w".., 7637)    = 7637
  29622:  close(18)                                       = 0
  29622:  fstat64(10, 0xFFBFF860)                         = 0
  29622:      Incurred fault #6, FLTBOUNDS  %pc = 0x00061500
  29622:        siginfo: SIGSEGV SEGV_MAPERR addr=0x0A453864
  29622:      Received signal #11, SIGSEGV [default]
  29622:        siginfo: SIGSEGV SEGV_MAPERR addr=0x0A453864

According to the above post and others, the problem may be related to
cnfs buffers...?  I'm using Solaris, with buffers > 2GB.  I did include
--enable-largefiles when configuring and didn't see any problems on
that front.  My cycbuff.conf looks like:

  cycbuff:NEWS01:/var/news/spool/cycbuffs/news01:4194304
  ...

and my buffers are:

  -rw-rw-r--   1 news     news     4294967296 Oct 17 18:06 news01
  ...

I could use < 2GB buffers, but I'm curious as to why this is happening
like this.  I seem to recall having very similar problems the last time
I set up INN (I don't know that I troubleshot it as far), and ended up
using 1.5GB buffers.

Thanks,
--nate

--------
nathan coraor
cse it support
nate at cse.psu.edu


More information about the inn-workers mailing list