Hello,

I seem to have encountered an efficiency bug.  Can anyone suggest ways
to investigate the cause?

When I say darcs pull in a certain project it says there are 260
patches, and if I say "a" for all, my cpu gets pegged for a long time.
 I've been doing darcs pull,
then accepting a few at a time, but it'll take forever that way.

I ran darcs pull -a overnight to see what happens, and it is still
running a large fraction (95% according to top) of cpu.  The memory
consumption seems relatively flat at just under 60 megs.

For now I'll leave it running, just for good measure.

Thanks,
Nathan Wilcox

Here's an update:

On 10/18/07, Nathan <nejucomo@gmail.com> wrote:
> Hello,
>
> I seem to have encountered an efficiency bug.  Can anyone suggest ways
> to investigate the cause?
>

...

> I ran darcs pull -a overnight to see what happens, and it is still
> running a large fraction (95% according to top) of cpu.  The memory
> consumption seems relatively flat at just under 60 megs.
>
> For now I'll leave it running, just for good measure.
>

It succeeded after a long while with exit status 0!  Here is the output:

$ time darcs pull -a
Pulling from "http://allmydata.org/source/tahoe/trunk"...
We have conflicts in the following files:
./src/allmydata/client.py ./src/allmydata/node.py
./src/allmydata/test/check_memory.py
./src/allmydata/test/test_runner.py
Warning: Not deleting ./src/foolscap/foolscap/slicers because it is non-empty.
Warning: Not deleting ./src/foolscap/foolscap because it is non-empty.
Warning: Not deleting ./src/foolscap because it is non-empty.
Warning: Not deleting ./src/simplejson/ez_setup because it is non-empty.
Warning: Not deleting ./src/simplejson because it is non-empty.
Warning: Not deleting ./src/zfec because it is non-empty.
Finished pulling and applying.

real    665m14.215s
user    633m39.716s
sys     1m41.388s

I'm not sure what triggered this, but I've just implemented a number of
improvements for darcs pull in the darcs 2 branch, one of which changes the
scaling properties.  Darcs 2 is not always faster on pull than darcs 1, but
sometimes it is.

Any chance you can try this same pull with darcs 2 (actually, the latest version
from http://darcs.net/repos/darcs-unstable), and then try it again using a
hashed repository?

David

I'm marking this as "deferred" until we get more feedback about how the
performance compares with Darcs 2.

Hello Nathan, 

Have you had any better luck with pull performing better when you try darcs 2?

  Mark

Hi Mark,

Thanks for the feedback.  I haven't tried darcs 2 yet, because I
typically rely on Ubuntu packages, (which is currently darcs 1.0.9rc1
on my system).  Next time I have some slack time in a project, I'll
install darcs 2 and get back to you.

Better yet, I'll continue using darcs 1.0.9rc1 until I trigger that
long hang situation, and that will serve as a reminder to compare
darcs 2's behavior.

Regards,
Nathan

On Feb 16, 2008 10:53 AM, Mark Stosberg <bugs@darcs.net> wrote:
>
>
> Hello Nathan,
>
> Have you had any better luck with pull performing better when you try darcs 2?
>
>   Mark
>
> ----------
> assignedto:  -> nejucomo
> title: pull takes a very long time (possibly improved in Darcs2) -> pull via http takes a very long time (possibly improved in Darcs2)
>
> __________________________________
> Darcs bug tracker <bugs@darcs.net>
> <http://bugs.darcs.net/issue550>
> __________________________________
>

Hi Nathan,

I'm going to mark this one resolved with the darcs 2 release (I keep hearing of
better performance on the mailing lists).  Please file a new report if you run
into more slowness.