All these buildbot failures recently share these things in common:

- they're happening on Ubuntu Feisty systems

- they are spamd tests

- they involve a spamd not giving its "server pid: \d+" line, which
  indicates that it's ready to receive connections, in time (within
  something like 3 minutes)

- upon examination, the spamd log looks like this:


[6910] dbg: rules: __MO_OL_25340 merged duplicates: __MO_OL_4EEDB
__MO_OL_7533E
[6910] dbg: conf: finish parsing
[6910] dbg: plugin:
Mail::SpamAssassin::Plugin::ReplaceTags=HASH(0x89db6e4) implements
'finish_parsing_end', priority 0
[6910] dbg: replacetags: replacing tags
[6910] dbg: replacetags: done replacing tags
[6910] dbg: bayes: tie-ing to DB file R/O ./log/user_state/bayes_toks
[6910] dbg: bayes: tie-ing to DB file R/O ./log/user_state/bayes_seen
[eof]


  or


[13512] dbg: rules: running head tests; score so far=9.205
[13512] dbg: rules: compiled head tests
[13512] dbg: rules: running head_eval tests; score so far=9.205
[13512] dbg: locker: safe_lock: created 
./log/user_state/auto-whitelist.lock.vmsa.13512
[13512] dbg: locker: safe_lock: trying to get lock on 
./log/user_state/auto-whitelist with 0 retries
[13512] dbg: locker: safe_lock: link to ./log/user_state/auto-whitelist.lock: 
link ok
[13512] dbg: auto-whitelist: tie-ing to DB file of type DB_File R/W in 
./log/user_state/auto-whitelist


  in other words, it seems to be hanging during the compile_now()
  operation, while trying to access a Bayes DB, or an AWL DB.  

  two examples:
  http://buildbot.spamassassin.org:8010/t-feisty-561/builds/270/step-test/0
  http://buildbot.spamassassin.org:8010/t-feisty-561/builds/268/step-test/0


- it's *very* intermittent.

I initially thought it was a vmware-related issue, but t-feisty-561 is
a non-emulated Pentium M laptop.

Is anyone else seeing anything like this behaviour on Feisty?  I don't
think it's limited to the test suite -- it's just that our tests are
successfully finding it.

--j.

Reply via email to