Differences between revisions 3 and 4
Revision 3 as of 2006-01-31 19:47:53
Size: 4184
Editor: JustinMason
Comment:
Revision 4 as of 2006-03-20 15:40:09
Size: 4214
Editor: JustinMason
Comment: note make disttest state
Deletions are marked like this. Additions are marked like this.
Line 62: Line 62:
 * '''tarball''': what's put in distributed tarballs  * '''tarball''': what's put in distributed tarballs (by make dist, make disttest)

Rule Life Cycle

The life-cycle of a rule goes like this.

  • A rule starts off in a developer's sandbox. It may be an experimental rule ("tflags nopublish" or "T_" prefix on the name).
  • Alternatively, it may be a non-experimental, but still in-sandbox, testing rule.
  • The developer may decide to switch the rule back and forth between those two states.
  • Non-experimental rules' promotability is measured (see SaUpdateBackend).

  • If it's good enough, it's published to the "active set".
  • A good rule may be manually copied from the sandbox to the "rulesrc/core" directory.
  • Eventually, it stops being good enough, through the normal attrition process for antispam rules, and it stops meeting the promotion criteria.

List Of Rule States

Rules in sandbox:

  • experimental -- don't promote me. "T_" prefix / "tflags nopublish" implies this. These rules are compiled, by the "build/mkrules" compiler at "make" time, to "rules/70_sandbox.cf".

  • s_poor -- promotable, but not meeting promotion criteria. Compiled to "rules/70_sandbox.cf".

  • s_good -- promotable, and meeting criteria. Rules in this state are copied into the "active set". Compiled to "rules/72_active.cf".

Rules in core:

  • c_poor -- promotable, but not meeting promotion criteria. Compiled to "rules/70_inactive.cf".

  • c_good -- promotable, and meeting criteria. Rules in this state are copied into the "active set". Compiled to "rules/72_active.cf".

Deleted rules:

  • gone -- rule has been deleted. If a rule is in c_poor for "an extended period of time", it goes here. (Right now, this has to be done manually.)

Legacy:

  • code_tied -- rule is very dependent on the distributed Mail::SpamAssassin perl modules, and is therefore still distributed as part of the "rules" directory in the code package. These are not compiled at all by "build/mkrules", and are always distributed.

(History: [http://www.nabble.com/hackathon-notes-from-Sat-p1887702.html mailing list message])

State Transitions

The permitted transitions for those rule states, therefore, are as follows:

  • experimental <---> s_poor

  • experimental <---> s_good

  • s_poor <---> s_good

  • c_poor <---> c_good

  • c_poor -> gone

List Of Build States

Some rules are only used from certain build states. Here are the list of states that SpamAssassin goes through, or that rules are packaged as, during various parts of its build process.

  • builddir: "./spamassassin", or similar, run from inside build dir

  • make_test: "make test"

  • mass_check: MassCheck run from inside "masses" dir

  • bbtest: the testing buildbot

  • bbmass: the bbmass buildbot

  • nightly: the NightlyMassCheck

  • make_install: what's installed via "make install"

  • tarball: what's put in distributed tarballs (by make dist, make disttest)

  • sa_update: what's delivered via sa-update

Build States vs. Rule States Matrix

And here's the table listing what rules are usable, where. {OK} indicates that a rule in that state is indeed usable from the listed build state.

experimental

s_poor

s_good

c_poor

c_good

code_tied

builddir

{OK}

{OK}

{OK}

{OK}

{OK}

{OK}

make_test

{OK}

{OK}

{OK}

{OK}

{OK}

{OK}

mass_check

{OK}

{OK}

{OK}

{OK}

{OK}

{OK}

bbtest

{OK}

{OK}

{OK}

{OK}

{OK}

{OK}

bbmass

{OK}

{OK}

{OK}

{OK}

{OK}

{OK}

nightly

{OK}

{OK}

{OK}

{OK}

{OK}

{OK}

make_install

{OK}

{OK}

{OK}

tarball

{OK}

{OK}

{OK}

sa_update

{OK}

{OK}

{OK}

RuleLifeCycle (last edited 2009-09-20 23:16:34 by localhost)