Hello, Kjartan.
You wrote in <mid:[EMAIL PROTECTED]>
ANV>> It's sounds strange... :(. But my mail corpus is more humble - I have
ANV>> only 700 letters of spam and 1000 of non-spam. My regard.rbd contains
ANV>> about 20000 key tokens and takes around 400KB. I tried it on Windows
ANV>> XP and Windows 98. In both cases all was o'key.
K�> I did build a new regard file with ~200 spam mails and ~200 non spam
K�> mails. The new regard.rdb file is around 100KB in size.
K�> This gave the same result as before. The Bat! started to consume more
K�> and more memory until i killed it.
K�> Then I tried both of the new builds of the plugin you gave me (the one
K�> w/o writing to log, and the other one which wrote alot of information
K�> to the log file).
K�> Both of the new builds failed the same way.
K�> Another observation is that there is no log file created, so the
K�> plugin must be failing before it tries to write the log file.
No, it can't. The log file is checked only when you finish your work
in The Bat! In no-log version it does'nt exist at all. If you say that
it doesn't work in other cases than it can't be a problem with
logfile.
My only guess is a bug which I described in describing the learning
machine - with quantity of tokens. Because of it you can build wrong
file with "negative" quantity - which regarded as very-very big
positive number. Then TBP_Initialize will try to read this huge number
of tokens...
Of course I corrected this bug and you can try to build regard.rbd
again by updated:
http://klirik.narod.ru/arc/baesyan.exe (286KB)
If it doesn't help, I want to ask you two things.
At first, when exactly the problem arises? Variants is:
1. It happens when you try to "install" the plugin into The Bat! In
this case The Bat! calls the critical functions of the plugin in next order:
1. TBP_Initialize (critical - opens and reads regard.rbd)
2. TBP_GetStatus (non-critical)
3. TBP_NeedConfig (non-critical - just stub always returned 0)
if NeedConfig returns wrong (-1) then
3a. TBP_Setup (non-critical - never called because of TBP_NeedConfig)
4. Two times TBP_GetName (non-critical)
5. Two times TBP_GetVersion (non-critical)
So, if problem arises when you only try to "install" the plugin than
the only place where it can occur is TBP_Initialize. It can be because
of wrong regard.rbd. In this case I need you regard.rbd to "touch".
Can you, please, send it to me?
2. Plugin is installed correctly and show himself in The Bat!. The
problem arises when you receive a letter - and real filtering actually
needed. In this case the critical section is TBP_GetSpamScore.
3. The problem arises when exiting The Bat!. Theoretically it can be
because of log. "Theoretically" - because if opening of logfile is
fail then the plugin doesn't try to write into it.
In attachment - real log created when I receive this your letter and
one spam.
K�> My system is Windows XP with all service packs applied and with all,
K�> except the very latest, critical updates applied.
K�> Best regards,
K�> Kjartan
--
Sincerely,
Alexey.
Using TB 1.63b7 on WinXP SP1 Corp + MUI RU, spelling by ORFO2002
mailto:[EMAIL PROTECTED]
***************************************************
N E W R E C O R D
***************************************************
Initialization completed.
NeedConfig called and returned 0.
1001: TBP_GetSpamScore called.. OK.
1001: Checking regarding table.. OK.
1001: Getting raw body message from The Bat!.. OK.
1001: Creating an instanse of frequency dictionary.. OK.
1001: Creatin an email parser instance for parsing the letter.. OK.
1001: Asigning the letter to emal parser.. OK.
1001: Parsing the letter.. OK.
1001: Receiving user info about the letter.. OK.
<[EMAIL PROTECTED]>
1001: Destroying email parser.. OK.
1001: Freeing processed raw email.. OK.
1001: Initializing the filter for 15 most interesting tokens.. OK.
1001: Regarding parsed frequency dictionary.. OK.
1001: Regarded value: 0.99
1001: Value for The Bat!: 99
1001: : -------------------------------------------
1001: services: 0.99
1001: request: 0.01
1001: private: 0.01
1001: performance: 0.99
1001: notified: 0.99
1001: information!: 0.99
1001: affiliate: 0.99
1001: _H Windows-1252: 0.99
1001: _H +0400: 0.01
1001: Submit: 0.01
1001: Special: 0.99
1001: Instructions: 0.99
1001: Help: 0.01
1001: Edition: 0.01
1001: AntiVirus: 0.01
1001: : -------------------------------------------
1001: Destroying used frequency dictionary.. OK.
1001: TBP_GetSpamScore successfully finished.
1002: TBP_GetSpamScore called.. OK.
1002: Checking regarding table.. OK.
1002: Getting raw body message from The Bat!.. OK.
1002: Creating an instanse of frequency dictionary.. OK.
1002: Creatin an email parser instance for parsing the letter.. OK.
1002: Asigning the letter to emal parser.. OK.
1002: Parsing the letter.. OK.
1002: Receiving user info about the letter.. OK.
<[EMAIL PROTECTED]>
1002: Destroying email parser.. OK.
1002: Freeing processed raw email.. OK.
1002: Initializing the filter for 15 most interesting tokens.. OK.
1002: Regarding parsed frequency dictionary.. OK.
1002: Regarded value: 1.16271e-030
1002: Value for The Bat!: 0
1002: : -------------------------------------------
1002: _H filter: 0.01
1002: _H errors-to: 0.01
1002: _H com?subject: 0.01
1002: _H austin: 0.01
1002: _H Kjartan: 0.01
1002: _H ISO-8859-15: 0.01
1002: _H <tbdev: 0.01
1002: _H <mailto: 0.01
1002: _H <http: 0.01
1002: _H 62.80.28.28: 0.01
1002: _H 2.1.1: 0.01
1002: _H +0400: 0.01
1002: TBUDLInfo: 0.01
1002: TBDEV: 0.01
1002: Kjartan: 0.01
1002: : -------------------------------------------
1002: Destroying used frequency dictionary.. OK.
1002: TBP_GetSpamScore successfully finished.
Deleting regarding base.. OK.
Finalization.
________________________________________________
Current version is 1.62 | "Using TBDEV" information:
http://www.silverstones.com/thebat/TBUDLInfo.html