Re: problem with trainin since upgrade to blamethesegfault

On Thu, Apr 03, 2008 at 04:30:59PM +1100, Jason Lewis wrote:
...
> > fprintf-->snprintf and placing thus generated message inside the
> > fatalerror() call, instead of pushing it out to stderr immediately.
> 
> Which error should be gone for good?
> 
> I still get the training error:
> 
> /tmp/crm114-20070810-BlameTheSegfault.src/crm114 -u 
> /home/jason/.crm114working/ /home/jason/.crm114working/mailreaver.crm 
> --good  < 
> /home/jason/Maildir/.ygood/cur/1207200462.M173503P6439V000000000000FD02I000000002C041267_0.debian,S=1353:2,S 
>   >  /home/jason/.crm114working/164561207200481989640077
> ifn='reaver_cache/known_good/reaver_cache/known_good/20080403_162729_032893_FFFFFFFFEDDBA61B', 
> filename=':*:gooddir::*:filename: 0 :*:decision_length:', errno=2, 
> error='No such file or directory'

either you didn't patch the .c, didn't recompile or are still running the old
binary. Modified src would have printed 
'For some reason, I was unable to read-open ...'
along with 'ifn=...'.

Anyway, I think Gerrit is right, there's no bug, the bugger is ... your 
patch :) - without the on purpose lines in mail*.crm we added on other
occasion, we're missing the context from the log and are faked to to believe
there's a bug.

Hum, all this happened around April 1, maybe there's a pttern ;)

But that's been useful to expose a bug (imo), here's the problem/dangerous
code:

#--[mailtrainer.crm]---
...
@462        {        #   Maybe it's a qualified name, maybe it's not
                input [:*:filename: 0 :*:decision_length:]
                trap /unable to read-open/
                output /\n COULDN'T READ THE GOOD FILE ':*:filename:' \n/
                alter (:_dw:) /:*:_nl:/
            }
...
alius
...
@499        {        #   Maybe it's a qualified name, maybe it's not
                input [:*:gooddir::*:filename: 0 :*:decision_length:]
                trap /unable to read-open/
                input [:*:filename: 0 :*:decision_length:]
                trap /unable to read-open/
                output /\n COULDN'T READ THE GOOD FILE ':*:filename:'\n/
                alter (:_dw:) /:*:_nl:/
            }
...
#--------------------------

our trouble line is @500, but I think that's the good one, missing from 
block above ALIUS, while the next INPUT try should be removed.
Both mailreaver and mailtrainer read the .cf where :gooddir: is defined, 
but :filename: passed by mailreaver includes :gooddir: as well.

Insted mailreaver should use just the bare filename, and mailtrainer 
shouldn't try-open here and there: it's supposed to work in a well defined
dir and read good/spam from well defined dirs, if it fails there's an error
or the file disappeared. 

Keeping the lines with :*dir: in makes sense since mailtrainer may run 
standalone.

Going to check mail{trainer,reaver}.crm, will post tonight hacked versions
if time permits.


-- 
 paolo
 
 GPG/PGP id:0x3A47DE45  - B5F9 AAA0 44BD 2B63 81E0  971F C6C0 0B87 3A47 DE45
 - 9/11: the outrageous deception and ongoing coverup: http://911review.org -

-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace