1186379 - Firefox reports every Jetpack add-on as "making Nightly run slowly"

Status: RESOLVED DUPLICATE of bug 1191342

(Toolkit :: Performance Monitoring, defect)

Description

[Marco Castelluccio [:marco]]

Very lightweight addons (such as https://github.com/mykmelez/tcpsocketpup/) are reported as slowing down the browser.

Also addons that I never use during a browsing session are reported as being slow (for example, "Firefox OS 2.0 Simulator", "Valence").

Comment 1

[David Teller [:Yoric] - still alive but not very active]

What does about:performance tell you about why the add-ons are reported?

Comment 2

[Marco Castelluccio [:marco]]

Attachment: about:performance screenshot

Comment 3

[David Teller [:Yoric] - still alive but not very active]

According to about:performance, Firefox OS 2.0 simulator has been activated during the session, and not just during startup. I'm not sure why, though. I'm waiting until bug 1157009 has been reviewed and landed, this will make things easier to understand.

We have a few sources of false alerts (see bug 1166889), but so far, I haven't had any reason to believe that add-ons that are never activated show up in the list by error.

Apparently, the Gecko Profiler was loaded. Did it tell you anything interesting?

Comment 4

[Marco Castelluccio [:marco]]

(In reply to David Rajchenbach-Teller [:Yoric] (use "needinfo") from comment #3)
> According to about:performance, Firefox OS 2.0 simulator has been activated
> during the session, and not just during startup. I'm not sure why, though.
> I'm waiting until bug 1157009 has been reviewed and landed, this will make
> things easier to understand.

I'm pretty sure I've never used it during this session (actually it's been a few
days since I last started it). Maybe it's doing something anyway even if it isn't
started?
It's weird that "API Enabler" is reported as well, since it doesn't really do
much: https://github.com/mykmelez/tcpsocketpup/blob/master/index.js#L81

> Apparently, the Gecko Profiler was loaded. Did it tell you anything
> interesting?

It wasn't running.
It is reported as well, even if I'm not profiling.

Comment 5

[David Teller [:Yoric] - still alive but not very active]

Mmmh...
Given that the numbers for all add-ons seem related, perhaps the AddonManager (or FHR?) is calling the addon-registered callback for some obscure reason?

Comment 6

[David Teller [:Yoric] - still alive but not very active]

Could you check in the Gecko Profiler if, by any chance, there is AddonManager activity? Or FHR activity?

Comment 7

[Marco Castelluccio [:marco]]

Attachment: iI_IG8OI.7z

The only AddonManager activity I see is related to the slow addon notification.
I can't see any FHR activity.

I'm attaching the profile.

Comment 8

[David Teller [:Yoric] - still alive but not very active]

As a side-note, about:telemetry => Keyed Histograms => MISBEHAVING_ADDONS_* recapitulates the measures that lead to performance alerts.

Comment 9

[David Teller [:Yoric] - still alive but not very active]

I don't see any activity of your add-ons in the profile, either.
Just to be sure: you filed this with a Nightly 41. Do you have the same behavior with 42?

Comment 10

[Marco Castelluccio [:marco]]

On 42 the simulator is no longer reported.
The Gecko profiler and API enabler are still reported (nearly a report every 60 seconds), with the same number of "Alerts" and "Jank level alerts".

Comment 11

[Marco Castelluccio [:marco]]

Forget it, I'm seeing the simulator reported again (still never used).

Comment 12

[David Teller [:Yoric] - still alive but not very active]

I expect that bug 1181175 will improve accuracy.

Also, can you confirm that you are running on Linux? Which kernel do you have?

Comment 13

[David Teller [:Yoric] - still alive but not very active]

Can you confirm that all the add-ons that are displayed as slow are jetpack-based?

Comment 15

[Marco Castelluccio [:marco]]

(In reply to David Rajchenbach-Teller [:Yoric] (use "needinfo") from comment #12)
> I expect that bug 1181175 will improve accuracy.
> 
> Also, can you confirm that you are running on Linux? Which kernel do you
> have?

Yes, I'm using Fedora 22, kernel 4.1.3.

(In reply to David Rajchenbach-Teller [:Yoric] (use "needinfo") from comment #13)
> Can you confirm that all the add-ons that are displayed as slow are
> jetpack-based?

I'm not sure about Valence.
tcpsocketpup and the simulators use the SDK (CCing Myk for confirmation).
The Gecko Profiler uses the SDK.

Comment 16

[Myk Melez [:myk] [@mykmelez]]

(In reply to Marco Castelluccio [:marco] from comment #15)
> (In reply to David Rajchenbach-Teller [:Yoric] (use "needinfo") from comment
> #13)
> > Can you confirm that all the add-ons that are displayed as slow are
> > jetpack-based?
> 
> I'm not sure about Valence.
> tcpsocketpup and the simulators use the SDK (CCing Myk for confirmation).

tcpsocketpup does for sure.  NI jryans to confirm that the simulators use it.

Comment 17

[J. Ryan Stinnett [:jryans] (Use needinfo, replies may be slow)]

(In reply to Myk Melez [:myk] [@mykmelez] from comment #16)
> (In reply to Marco Castelluccio [:marco] from comment #15)
> > (In reply to David Rajchenbach-Teller [:Yoric] (use "needinfo") from comment
> > #13)
> > > Can you confirm that all the add-ons that are displayed as slow are
> > > jetpack-based?
> > 
> > I'm not sure about Valence.
> > tcpsocketpup and the simulators use the SDK (CCing Myk for confirmation).
> 
> tcpsocketpup does for sure.  NI jryans to confirm that the simulators use it.

I am not sure what form of "using" the SDK is relevant here, so I will try to be as precise as possible.

The simulators use a custom bootstrap.js[1] (not created by SDK or JPM packaging tools), which then invokes the SDK CommonJS loader machinery, similar to most SDK based add-ons.

Valence uses the standard JPM bootstrap.js[2], which effectively delegates its work to a Firefox-controlled bootstrap.js[3].  That file then also invokes the standard SDK loader machinery as well.

[1]: https://dxr.allizom.org/mozilla-central/source/b2g/simulator/bootstrap.js
[2]: https://github.com/mozilla/valence/blob/master/template/bootstrap.js
[3]: https://dxr.allizom.org/mozilla-central/source/addon-sdk/source/lib/sdk/addon/bootstrap.js

Comment 18

[David Teller [:Yoric] - still alive but not very active]

This seems to be the same as bug 1191342, which itself seems to be a problem in Jetpack itself.