Re: [PATCH 0/5] Use watchman to reduce index refresh time

[Christian Couder <christian.couder@gmail.com>]

On Tue, Nov 10, 2015 at 10:04 PM, David Turner <dturner@twopensource.com> wrote:
> On Mon, 2015-11-09 at 21:06 +0100, Christian Couder wrote:
>> Using David's series I get worse results than all of the above but I
>> guess it's because his series is based on an ancient git version
>> (v2.0.0-rc0).
>
> My more-recent series is on top of 2.4, but (for webkit):
> mine: 0m0.206s
> duy's: 0m0.107s

I tried using your more recent series and I get basically no change
compared to the current git (without using untracked cache) on Webkit
with around 5600 untracked files. Maybe I am doing something wrong.

I compiled watchman and installed it. It is in /usr/local/bin/watchman
but I am not sure it is actually being used though I have configured
"core.usewatchman" to "true".

> However, I'm getting occasional index-helper segfaults due to
> istate->last_update being NULL.  (I'm using Duy's  index-helper branch
> from his github + this patchset + a function signature fix due to a
> newer version of libwatchman).  I haven't looked into why this is --
> maybe accidentally mixing git versions while testing?
>
> Also, after messing around for a while (on Duy's branch), I ended up in
> a state where git status would take ~2.5s every time.  The index helper
> was alive but evidently not working right.  Killing and restarting it
> worked.  Actually, I think I can repro this more easily: "git rm
> Changelog" seems to put the index-helper into this state.

Thanks for this. I have also seen strange things happening with the
index-helper.
When it is working I found that it provides around 10% improvement on
git rebase time.