[Bug 6855] Rules with extremely poor CPU_time / mail_size ratio

[bu...@spamassassin.apache.org]

https://bz.apache.org/SpamAssassin/show_bug.cgi?id=6855

Henrik Krohns <he...@hege.li> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hege@hege.li

--- Comment #12 from Henrik Krohns <he...@hege.li> ---
__FILL* monsters still dominating runtimes..

T        __FILL_THIS_FORM_SHORT2  118.8653    0.2009 24884
T         __FILL_THIS_FORM_LONG2  111.5167    0.1723 24884
T                __E_LIKE_LETTER   48.7511    0.0261 24884
T  __FILL_THIS_FORM_FRAUD_PHISH1   48.7025    0.1394 24884
T                      __LOWER_E   44.6741    0.0277 24884
T         __FILL_THIS_FORM_LOAN1   35.8043    0.0469 24884
T            HTTPS_HTTP_MISMATCH   33.9572    1.1763 24884
T                    __MY_VICTIM   22.8257    0.0293 24884
T        __FILL_THIS_FORM_SHORT1   18.6945    0.0259 24884
T             __UPPERCASE_75_100   18.6015    0.1381 24884
T                    FUZZY_ERECT   18.1948    0.0405 24884
T         __FILL_THIS_FORM_LONG1   17.5074    0.0392 24884

Problem is they seem to work very well, but look impossible to rewrite better,
there's gazillion optional matches and alternations going on.

Is there _any_ simple common identifier in these mails, word, phrase lists?

Perhaps for 4.0.0 we could introduce some meta/tflags option to run specified
rules only if some other rule hit first, this would work by adjusting
priorities.

In theory we wouldn't even need any new options, but I don't know how much code
it would require to handle metas intelligently and skip running any subrules
that pending metas do not require any more.

-- 
You are receiving this mail because:
You are the assignee for the bug.