Bug Tracker item #3396536, was opened at 2011-08-22 14:45 Message generated for change (Comment added) made by jguidini You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=1126467&aid=3396536&group_id=250683
Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: daemon Group: v3.10.x Status: Open Resolution: None Priority: 5 Private: No Submitted By: Juliano Alves Guidini (jguidini) Assigned to: Nobody/Anonymous (nobody) Summary: Daemon trap divide error Initial Comment: DSPAM 3.10.1 in daemon mode works great for 5, 6 days ( about 380 MB hash), when it crash. See syslog: Aug 22 09:38:59 fspam kernel: [867638.233088] dspam[26311] trap divide error ip:f77c76eb sp:f19de660 error:0 in libdspam.so.7.0.0[f77b1000+19000] Compiled with this options: Configuration parameters: '--enable-daemon' '--enable-virtual-users' '--enable-debug' '--enable-verbose-debug' '--enable-large-scale' 'CFLAGS=-mtune=i386 -m32' System: Linux fspam 2.6.32-5-amd64 #1 SMP Tue Jun 14 09:42:28 UTC 2011 x86_64 GNU/Linux ( Debian 6 ) I try compile dspam whit CFLAGS= -m32, but not solve. Thanks! ---------------------------------------------------------------------- >Comment By: Juliano Alves Guidini (jguidini) Date: 2011-08-24 09:44 Message: I think in eliminate memory leak, array length explosion.. exception in float point etc.. because de trap error (divide by zero). with im -m32, -mtune=i386 remove all performance/optimization of CPU architecture (Xeon) with a pure i386 co dspam --version DSPAM Anti-Spam Suite 3.10.1 (agent/library) Copyright (C) 2002-2011 DSPAM Project http://dspam.sourceforge.net. DSPAM may be copied only under the terms of the GNU Affero General Public License, a copy of which can be found with the DSPAM distribution kit. Configuration parameters: '--enable-daemon' '--enable-virtual-users' '--enable-debug' '--enable-verbose-debug' '--enable-large-scale' 'CFLAGS=-ggdb' ---------------------------------------------------------------------- Comment By: Tom Hendrikx (whyscream) Date: 2011-08-24 09:40 Message: I'm no complete gcc guru, but it seems to me that both -mtune=i386 and -m32 are trying to accomplish the same thing: creating 32bit binaries. Since your host machine is 64bit, why are you doing this? ---------------------------------------------------------------------- Comment By: Juliano Alves Guidini (jguidini) Date: 2011-08-24 09:39 Message: Recompiled! Waiting a crash... ---------------------------------------------------------------------- Comment By: Juliano Alves Guidini (jguidini) Date: 2011-08-24 09:18 Message: I will recompile. About -m32 initially was compiled without -mtune=i386 -m32, and crash. I remove de hash and work well a while. And crash... I recompile -mtune=i386 -m32 for testing, and crash again. Now I will recompile without -mtune=i386 -m32 and with -ggdb and remove de hash. Hash 32 bits not was touched by hash 64 bits, every recompiled, new hash (deleted the old). ---------------------------------------------------------------------- Comment By: Tom Hendrikx (whyscream) Date: 2011-08-23 18:24 Message: The backtrace is missing some data (no debugging symbols found). Is it possible that you recompile dspam with CFLAGS="-ggdb", and make sure that you don't strip the binaries? Also I see that you are trying to compile 32bit code (-m32) on a 64-bit machine? Any special reason for this? Finally: could it be that the hash is also touched by a 64bit dspam? This is not supported. For above test it would be best to start with a clean data dir (i.e. don't re-use the existing hash data), so that we can be sure that the issue is with the dspam code and not with the hash data. ---------------------------------------------------------------------- Comment By: Juliano Alves Guidini (jguidini) Date: 2011-08-23 11:57 Message: Hi, output to this bug attached, thank! ---------------------------------------------------------------------- Comment By: Tom Hendrikx (whyscream) Date: 2011-08-23 11:13 Message: Hi, you'll need to analyse the coredump until you have a backtrace. Please do the following on teh machine that generated the coredump: $ gdb /usr/bin/dspam --core <coredump> > thread apply all bt full > quit Then, attach the output to this bug, or send it to the devel- list by e-mail. ---------------------------------------------------------------------- Comment By: Robson Eisinger (papilobr) Date: 2011-08-23 10:33 Message: Hi guys, I work with jguidini, we got a coredump with 83Mb, so we need a way to share it. Or is possible to reduce it?To be honest this is our first time generating a coredump. Any advice is very welcome. Cheers, ---------------------------------------------------------------------- Comment By: Juliano Alves Guidini (jguidini) Date: 2011-08-23 10:12 Message: Whyscream, I reproduce a core dump with 3.10.1, but is a big file... 83 MB.. attach file? ---------------------------------------------------------------------- Comment By: Juliano Alves Guidini (jguidini) Date: 2011-08-23 10:01 Message: Hi, File attached. dspam --version DSPAM Anti-Spam Suite 3.10.1 (agent/library) Copyright (C) 2002-2011 DSPAM Project http://dspam.sourceforge.net. DSPAM may be copied only under the terms of the GNU Affero General Public License, a copy of which can be found with the DSPAM distribution kit. Configuration parameters: '--enable-daemon' '--enable-virtual-users' '--enable-debug' '--enable-verbose-debug' '--enable-large-scale' 'CFLAGS=-mtune=i386 -m32' -gcc version 4.4.5 (Debian 4.4.5-8) -automake (GNU automake) 1.11.1 -autoconf (GNU Autoconf) 2.67 I get the package from where ( https://sourceforge.net/projects/dspam/files/dspam/ ) and compile it. Thanks! ---------------------------------------------------------------------- Comment By: Paul Cockings (paulcockings) Date: 2011-08-22 18:54 Message: Can you attached your dspam.conf to this bugID? (excluding any usernames/passwords). I'm interested to see your settings for Algorithm, Tokenizer, PValue etc Maybe also post the full unmodified output from # dspam --version Can I also ask if your are installing from source or using a package (if so, where is the package from) Also... can you advise on your versions of gcc, automake, autoconf Many thanks, ---------------------------------------------------------------------- Comment By: Tom Hendrikx (whyscream) Date: 2011-08-22 17:11 Message: Can you manage to get a coredump after a crash? We could analyse that in order to find out where the error is. If you don't have one and need to reproduce, then please do so with a 3.10.1 binary, not with an older one. ---------------------------------------------------------------------- Comment By: Juliano Alves Guidini (jguidini) Date: 2011-08-22 16:53 Message: Thanks for your fast reply! :-) First, I installed dspam 3.9.1-RC1, and after a while it crashed, So, I removed all the hash driver data and started it again. That lasted for a few days, but still crashed. I tried the stable version (3.9.0), with the same result. Now testing it with the version (3.10.0) and the lastest version (3.10.1) and nothing. In my last try, I recompiled it using the following options -mtune=i386 -m32 (3.10.X), and the bug remains. If you need any data to help solving this issue, just ask. Thanks again. ---------------------------------------------------------------------- Comment By: Paul Cockings (paulcockings) Date: 2011-08-22 16:30 Message: Hello, and thanks for reporting your issue. May I ask if you consider this to be a new bug - i.e. you have been using dspam before and this issue was not present. If you try the 3.9.0 release do you have the same issues? Many thanks, ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=1126467&aid=3396536&group_id=250683 ------------------------------------------------------------------------------ EMC VNX: the world's simplest storage, starting under $10K The only unified storage solution that offers unified management Up to 160% more powerful than alternatives and 25% more efficient. Guaranteed. http://p.sf.net/sfu/emc-vnx-dev2dev _______________________________________________ Dspam-devel mailing list Dspam-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dspam-devel
