[darcs-users] Darcs 2.4 - performance testing volunteers needed

[Simon Marlow]

On 22/12/09 17:12, Eric Kow wrote:

> So could I get a benchmarking volunteer for each of these platforms?
> It would be extremely helpful if you could respond by the end of this
> week :-)
>
>   - MacOS X
>   - Linux
>   - Windows

I've just built the latest darcs so that I could use --skip-conflicts 
(it's great!).

I'm doing all my work in hashed repos locally, and my perception of the 
performance is that it's "different" to darcs-1 format repos. Some 
things are very fast: whatsnew, pull/push --dry-run. Other things seem 
to crunch my disk a lot and take a lot longer than I would expect.  For 
example, I just pushed a few patches from one GHC repo to another and it 
went like this:

Shall I push this patch? (7/7)  [ynWvplxdaqjk], or ? for help: y
Writing inventory 3 done, 30 queued. 
0000330575-fa3eaa4bf6ddc10a086115a0b3ad2aa2
Writing inventory 4 done, 29 queued. 
0001980260-fa47963e3f314d52c3e1aa9136e00cd0
Writing inventory 33 done, 0 queued. 
0000218257-06c742c93d020a1c3bbac14af46a1c8e
Finished applying...
Push successful.

why is it writing *all* of the inventories?  This is perhaps related to 
http://bugs.darcs.net/issue1589. I thought the point of tagginng and 
optimising is to avoid needing to write the old inventories?

Cheers,
	Simon