Re: HEAD broken ...

[Justin Mason <jm...@jmason.org>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


Theo Van Dinter writes:
> On Thu, Feb 19, 2004 at 04:36:12PM -0600, Michael Parker wrote:
> > I had to do a make clean before it would start behaving.  I'm not
> > seeing any problems at the moment.
> 
> Well, I'm at the latest SVN release (r6781 right now), and a full regression:
> 
> test -f Makefile && make distclean ; perl Makefile.PL < /dev/null ; make ; make test
> 
> fails horribly:
> 
> PERL_DL_NONLAZY=1 /usr/local/bin/perl -Iblib/arch -Iblib/lib -I/usr/lib/perl5/5.6.1/i386-linux
> -I/usr/lib/perl5/5.6.1 -e 'use Test::Harness qw(&runtests $verbose); $verbose=0; runtests @ARGV;' t/*.t
> t/basic_lint........ok                                                       
> t/bayesdbm..........ok 12/43Can't call method "finish" on unblessed reference at
> ../blib/lib/Mail/SpamAssassin/MsgContainer.pm line 671.
> t/bayesdbm..........dubious                                                  
>         Test returned status 2 (wstat 512, 0x200)
> DIED. FAILED tests 13-43
>         Failed 31/43 tests, 27.91% okay
> t/bayessql..........skipped test on this platform
> t/cidrs.............ok                                                       
> [...  ^C hit to stop the failures  ...]

I'm pretty sure this is now fixed.  Could you recheck?  "make disttest"
now works for me.

> > I did notice fairly high memory usage on a mass-check run this
> > morning, but hadn't been paying attention previously to know for sure
> > if this was a problem.
> 
> For me, the nightly run was taking ~2 hours and ~40MB RAM.  I killed it
> this afternoon when I saw it was still running after 10 hours and had
> OOMed the box (2GB total VM) numerous times.
> 
> I wasn't thrilled...  JM said he fixed that this morning though, but I
> haven't had a chance to test.  Until then, I've disabled the nightly
> runs on my box.

Yeah, should be fixed.   there were circular references between
the message and metadata objects -- which means no GC'ing in perl-land.

> I'm still leaning towards either reverting out the metadata
> code, or disabling it in the code (probably difficult from a
> quick glance) until we finish coming up with the API we want for
> MsgContainer/MsgParser/MsgMetaData/etc.

noooo!!!!!  it works [*], it's valuable, it's something we needed
to get in for 3.0.0 (IMO).

(*: modulo some "teething pains". ;)

> > > <sigh>  Someone's buying drinks for the dev team. ;)
> > 
> > I'd be more than happy to, especially after yesterday. :)
> 
> Are you saying you broke HEAD? ;)

Damn, I deliberately never mentioned the traditional "who breaks HEAD
buys pints" rule, but seems it got around through other channels ;)

- --j.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)
Comment: Exmh CVS

iD8DBQFANUBwQTcbUG5Y7woRApSWAKCUXAFIzGmr+OCrPUdsluVs8RT2WwCfeHfv
c+jr+ukWT0vNyw74VEhDurc=
=Euj1
-----END PGP SIGNATURE-----

Re: HEAD broken ...

[Theo Van Dinter <fe...@kluge.net>]

On Thu, Feb 19, 2004 at 03:02:08PM -0800, Justin Mason wrote:
> (*: modulo some "teething pains". ;)

There's "teething pains" and "we need to revive pains"...

> Damn, I deliberately never mentioned the traditional "who breaks HEAD
> buys pints" rule, but seems it got around through other channels ;)

Heh.  I think it's pretty standard in SW development shops actually.
Either drinks or money in a jar or something.

-- 
Randomly Generated Tagline:
Perhaps I'm missing the gene for making enemies.  :-)
              -- Larry Wall in <19...@wall.org>