Re: [Chicken-users] long runtime of salmonella report

[Alan Post]

On Sun, Jan 09, 2011 at 12:31:10AM +0100, Felix wrote:
> From: Alan Post <address@hidden>
> Subject: [Chicken-users] long runtime of salmonella report
> Date: Sat, 8 Jan 2011 16:26:45 -0700
> 
> > For the last two days, the Salmonella report has taken a
> > signficantly longer time to run than it should.
> > 
> > The problem is the test suite for the jbogenturfa'i egg.
> > I'm tracking down a significant slowdown in the program,
> > where one test suite was running in 20.6 seconds on the
> > 6th and took 1524.6 and 1662.0 seconds over the next
> > two days, respectively.  As well, I added more tests,
> > all of which are seeing the same slowdown.
> > 
> > This egg takes 30-60 minutes for me to compile, which
> > makes my debug cycle frustratingly long.  I'm aware of
> > and working on the problem, and will either remove/disable
> > the tests or find and fix the problem over the next couple
> > of days.
> > 
> > If the runtime of the salmonella report is significantly
> > affecting you, please report so and I'll disable the
> > jbogenturfa'i tests today.  I would like to add some timing
> > tests and otherwise test using the testing system, so if
> > it isn't significantly affecting you, I'd appreciate your
> > patience.
> 
> Hi, Alan!
> 
> I don't know whether this is a problem (Mario?), but you could
> run certain tests only when (say) an environment variable is
> set, or use some other means of "parameterizing" the set of
> tests executed.
> 

Before I committed the changes, the whole test suite (6176 tests)
ran in 2-3 minutes on my machine.  While working on the program,
one time I compiled the code I saw this slowdown, but after
recompiling (in part to determine why it suddenly got slow) the
problem went away.

The salmonella report is the first time I'm seeing the problem in an
apparently stable way--it's run with the same slowness two days in a
row, which I have not seen until now.

Once I discover what is going on, I'd expect the test suite to run
in 2-3 minutes, rather than in the 15 hours (!) it is taking now.  I
haven't made the kind of change that can easily explain such an
explosion in the runtime--I've thought I was largely making cosmetic
changes.  At any rate, the problem isn't yet obvious to me.

Mario and I spoke yesterday, and I asked him to let me have a couple
days at least to experiment with.  Yesterday I didn't even know
whether the problem would repeat itself today--it certainly hasn't
been doing that as I've been compiling the code on my local machine.

-Alan
-- 
.i ko djuno fi le do sevzi