Are the patchbots down?

[Simon King]

Hi!

Is there currently a problem with the patchbots on trac? At #13589, we are
currently struggling against a 5% regression of startup time, hence,
we'd really need to see the results of the startup time plugin. But
alas, the last test of the patches happened 2013-06-20 00:41:11, even
though I kicked repeatedly the patchbot in the last few days.

Talking about the plugin: Is there an easy way to use the plugin on my
machine, so that there'd be no need to wait for the patchbots on trac?

Talking about speed regression: Probably Nicolas can say more about
it. But to my understanding, #13589 is a building block of Nicolas'
attempt to make the category framework better scalable. The fine tuning
of the "web of categories" is currently done manually, which will soon
become a maintenance nightmare when trying to extend the scope of Sage's
category framework. Nicolas makes major parts of it automatic.

Is this worth paying 5% of startup time, which is about 75 ms on my
machine?

I'd say "yes", but of course I do my best to try and prevent any regression.

Best regards,
Simon

[Dima Pasechnik]

On 2013-06-27, Simon King <simon...@uni-jena.de> wrote:
> Hi!
>
> Is there currently a problem with the patchbots on trac?

arando patchbot is down, as it is a box in my office, and after work on
power grid it failed to come back online properly.
So it's probably will take me to fix it face-to-face, something that's
not going to happen before 2nd week of August. Sorry.

Dima

[Volker Braun]

Same here, my desktop dropped off the net and will probably have to wait until I'm back home.

[Christian Nassau]

I've just restarted my patchbot, but I don't know how long it will
remain up since that machine (running OpenSuse 12.3) currently suffers
from pretty regular kernel oopses.

By the way, about half the time the patchbot reports a startup-speed
regression for the core package and refuses to test other tickets. I
wonder if that makes sense? After all, what's the point of reference for
the base package itself?

Cheers,
Christian

Am 28.06.2013 03:24, schrieb Volker Braun:
> Same here, my desktop dropped off the net and will probably have to wait
> until I'm back home.
>
>
> On Thursday, June 27, 2013 5:40:48 PM UTC-4, Dima Pasechnik wrote:
>

> On 2013-06-27, Simon King <simon...@uni-jena.de <javascript:>> wrote:
> > Hi!
> >
> > Is there currently a problem with the patchbots on trac?
>
> arando patchbot is down, as it is a box in my office, and after work on
> power grid it failed to come back online properly.
> So it's probably will take me to fix it face-to-face, something that's
> not going to happen before 2nd week of August. Sorry.
>
> Dima
>
>

> --
> You received this message because you are subscribed to the Google
> Groups "sage-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to sage-devel+...@googlegroups.com.
> To post to this group, send email to sage-...@googlegroups.com.
> Visit this group at http://groups.google.com/group/sage-devel.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

[Simon King]

Hi!

On 2013-06-27, Simon King <simon...@uni-jena.de> wrote:
> Is there currently a problem with the patchbots on trac?

According to your answers, indeed some bots went down. But now, the
colourful blobs show up again.

> At #13589, we are
> currently struggling against a 5% regression of startup time

... and: With the new patches from #13589, the plugin now even finds a
non-significant decrease in startup time (in particular, no regression).

At this point, thank you for those who implemented the plugin.
Otherwise, I would not have bothered to try and make it faster. We are
talking about functions that are called about 100 times during startup,
but hardly ever *after* Sage has started. Only the plugin showed that it
would make sense to optimize them.

> Talking about the plugin: Is there an easy way to use the plugin on my
> machine, so that there'd be no need to wait for the patchbots on trac?

Is there?

Best regards,
Simon

[Volker Braun]

I had my desktop rebooted manually.

You can't call the plugin directly, but it is very straightforward code and it would be easy to turn it into a standalone script that just compares any two sage trees.

[Simon King]

Hi Volker,

On 2013-06-28, Volker Braun <vbrau...@gmail.com> wrote:
> You can't call the plugin directly, but it is very straightforward code and
> it would be easy to turn it into a standalone script that just compares any
> two sage trees.

This would be nice to have! As I wrote on the other thread, I took 20
samples of "sage -startuptime" without and 20 with the patches, but I
am not totally confident how to interprete it.

Best regards,
Simon

[Volker Braun]

On Friday, June 28, 2013 10:56:48 AM UTC-4, Simon King wrote:

This would be nice to have! As I wrote on the other thread, I took 20
samples of "sage -startuptime" without and 20 with the patches, but I
am not totally confident how to interprete it.

Yes, especially since it turns out to be not to approximate a normal distribution particularly well. The patchbot does a non-parametric test: http://en.wikipedia.org/wiki/Mann%E2%80%93Whitney_U

[Simon King]

Hi Volker,

On 2013-06-28, Volker Braun <vbrau...@gmail.com> wrote:

> Yes, especially since it turns out to be not to approximate a normal
> distribution particularly well. The patchbot does a non-parametric test:
> http://en.wikipedia.org/wiki/Mann%E2%80%93Whitney_U

Thanks!

[Robert Bradshaw]

On Fri, Jun 28, 2013 at 5:25 AM, Simon King <simon...@uni-jena.de> wrote:
> Hi!
>
> On 2013-06-27, Simon King <simon...@uni-jena.de> wrote:
>> Is there currently a problem with the patchbots on trac?
>
> According to your answers, indeed some bots went down. But now, the
> colourful blobs show up again.
>
>> At #13589, we are
>> currently struggling against a 5% regression of startup time
>
> ... and: With the new patches from #13589, the plugin now even finds a
> non-significant decrease in startup time (in particular, no regression).

Nice work.

> At this point, thank you for those who implemented the plugin.
> Otherwise, I would not have bothered to try and make it faster. We are
> talking about functions that are called about 100 times during startup,
> but hardly ever *after* Sage has started. Only the plugin showed that it
> would make sense to optimize them.

Note that Sage startup is so noisy that it's really hard to gather
good information on it. As mentioned, it does use the Mann-Whitney U
test, but it's still a pretty small sample size, but it's still a very
useful signal. IMHO, Sage's startup time is death by a thousand cuts,
1% here, 5% here, it really compounds, and people just aren't aware.
(The number of imported modules just keeps creeping up as well... but
often it's just being aware and saying "is that worth it or should it
be worked around?")

Now that we're (mostly) out of the dark ages with respect to
continuous testing, I can't wait to get out of the dark ages with our
revision control system :).

>> Talking about the plugin: Is there an easy way to use the plugin on my
>> machine, so that there'd be no need to wait for the patchbots on trac?
>
> Is there?

You can run the patchbot on a specific ticket and in plugin-only mode.
It'd be good to be able to specify a specific plugin (though many of
them compare to the base branch, so that'd be harder to do out of
context).

Even better is to run a patchbot yourself. (If you want you can set
very high priority to tickets authored by yourself.)

- Robert