This has been reported by at least two people (me and Jason, although
Jason seems to have noticed it only for benchmarking).

We don't yet have any idea why this would be the case.

So we basically just need a lot of diagnostic actions and/or thinking to
rule out possibilities.

Some thoughts may be:
- Some sort of HFS+ particularities, eg. wrt hardlinks?
- Spotlight
- Time Machine (off for benchmarking, though)

What could be causing this? And how do we confirm the hypothesis?
Even darcs get on the local filesystem is slow, mind you...

Eric Kow writes:

 > Some thoughts may be:
 > - Some sort of HFS+ particularities, eg. wrt hardlinks?

Maybe.

 > - Spotlight

Shouldn't be; that runs in the background.

 > - Time Machine (off for benchmarking, though)

Nope.  It's off, period, for me (everything I do is under VC).

 > What could be causing this? And how do we confirm the hypothesis?
 > Even darcs get on the local filesystem is slow, mind you...

All system calls on Mac are slow, that's all there is to it AFAICS.

Florian's dtrace results (attached) look particular important to
snapshot on this ticket.

http://lists.osuosl.org/pipermail/darcs-users/2010-July/024655.html

96% of our time spent in __semwait_signal, which actually isn't called
all that much... what does this mean?

Attachments

• florian-dtrace

There's a hypothesis out there about stat() being related to this