[darcs-users] darcs-benchmark 0.1.9 - benchmarking help wanted for 2.5

[Eric Kow]

On Tue, Aug 17, 2010 at 09:52:25 -0400, Nathan Gray wrote:
> > Perhaps tar-bzip2 your ~/.darcs-benchmark directory and send it
> > our way.  It could provide some clues, although I'm a bit doubtful
> 
> Attached.

Hmm, I notice that you have some darcs-2 format timings for the
"cap - darcs-2 format" repository in the 2010-08-13 timings file.

> I tried several.  I think the problem is that I tried to do a
> bunch of them all together over the weekend.  When that froze, or
> died, or whatever it did when I was not watching, I am now not
> able to run benchmarks on any of those repos.
> 
>   $ darcs-benchmark run darcs-2.4.98.3 / cap-hashed
>   No benchmarks to run for cap - hashed format

Do you have configuration files for those repositories as well?

Also, I wonder if the filtering mechanism for repository variants
is related to this.  This is what it currently looks like (sorry for the
Haskell):

    -- | The subset of variants appropriate to the given darcs version
    appropriateVariants :: TestBinary -> [Variant] -> [Variant]
    appropriateVariants bin = filter (appropriate . vId)
      where appropriate GitVariant
              | vcs /= VCSGit = False
            appropriate HgVariant
              | vcs /= VCSHg = False
            appropriate OptimizePristineVariant
              | vcs /= VCSDarcs || ver < [2,3,97] = False
            appropriate DefaultVariant
              | vcs /= VCSDarcs || ver > [2,4,96] = False
            appropriate _ = True
            vcs = binVCS bin
            ver = binVersion bin
    
It's saying that we never use the OptimizePristineVariant variant before
darcs 2.3.97 and we never use the DefaultVariant after 2.4.96.  The latter
sounds like an interesting clue.  Do we have an optimize pristine variant
for the cap repo?

Now that this infrastructure is in place, you may be interested to note that we
may be able to put in logic of sort "if version < 2, only use the old-fashioned
variant" or "avoid using using old-fashioned variant in darcs >= 2"

> That seems to have worked.  Running on my cap repos still results
> in the same error.

does darcs-benchmark report say anything about cap?

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
For a faster response, try +44 (0)1273 64 2905 or
xmpp:kowey at jabber.fr (Jabber or Google Talk only)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.osuosl.org/pipermail/darcs-users/attachments/20100818/a2f096f0/attachment.pgp>