835584 - Precompile and preload some more scripts that are on the critical startup path

Status: RESOLVED FIXED

(Core :: General, defect)

Description

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

forms.js and UserAgentOverrides.jsm overhead was significant on startup.

Comment 1

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Attachment: Precompile forms.js and preload UserAgenOverrides.jsm

Comment 2

[[:fabrice] Fabrice Desré]

Comment on attachment 707338 [details] [diff] [review]
Precompile forms.js and preload UserAgenOverrides.jsm

Review of attachment 707338 [details] [diff] [review]:
-----------------------------------------------------------------

::: dom/ipc/preload.js
@@ +71,5 @@
>    Cc["@mozilla.org/thread-manager;1"].getService(Ci["nsIThreadManager"]);
>    Cc["@mozilla.org/toolkit/app-startup;1"].getService(Ci["nsIAppStartup"]);
>    Cc["@mozilla.org/uriloader;1"].getService(Ci["nsIURILoader"]);
>  
> +  Services.io.getProtocolHandler("app");

I'm not sure this is useful, since the AppsProtocolHandler constructor doesn't do much.

Comment 3

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

I added that so we execute AppProtocolHandler.js code, but I didn't test it in isolation so you're right that it may not be a win.

Comment 4

[Vivien Nicolas (:vingtetun) (:21) - (NOT reading bugmails, needinfo? please)]

(In reply to Chris Jones [:cjones] [:warhammer] from comment #3)
> I added that so we execute AppProtocolHandler.js code, but I didn't test it
> in isolation so you're right that it may not be a win.

It worth it imho. All the JS API / Modules instantiated on the main loading path seems to hurt.

Also are you not going to call UserAgentOverrides.init(); twice? Once in the preload process and once in uao_child.js when it will be inserted via loadFrameScript?

Comment 5

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

(In reply to Chris Jones [:cjones] [:warhammer] from comment #3)
> I added that so we execute AppProtocolHandler.js code, but I didn't test it
> in isolation so you're right that it may not be a win.

This gets 5-6 samples out of a profile in AppProtocolHandler.js:1.  I think this is a valid win.

Comment 6

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

(In reply to Vivien Nicolas (:vingtetun) (:21) from comment #4)
> Also are you not going to call UserAgentOverrides.init(); twice? Once in the
> preload process and once in uao_child.js when it will be inserted via
> loadFrameScript?

The code has an "already-initialized" check.  So we call it twice but only run it once.

Comment 7

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Oh, I see what you mean.  We only TryCacheAndCompile("UAO_child.js"), we don't run it.  But since it doesn't depend on content, we could go ahead and run that and remove the explicit call to UAO.init().

Comment 8

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Actually, with the way UAO_child.js is set up, we're technically breaking the rules here and always loading the overrides, regardless of product.  The way it's set up right now we can't legally preload it :|.

Comment 9

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Any suggestion for how we can do this without resorting to something "unclean" like a feature-pref for overrides?

Comment 10

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

For example, I might write code to check if there are any override prefs set and then load the UAO code if so, but that might horrify chrome hackers.

Comment 11

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Feedback ping.  This is a very nice measurable win.

Comment 12

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

I'm going to add a feature pref for UA overrides and check it in preload.js to see whether to load and init() UserAgentOverrides, and remove UAO_child.js entirely.

Anyone object?

Comment 13

[[:fabrice] Fabrice Desré]

I would be very happy to see UAO_child.js disappear.

Comment 14

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Attachment: Precompile forms.js and preload UserAgenOverrides.jsm, v2

Comment 15

[[:fabrice] Fabrice Desré]

Comment on attachment 710432 [details] [diff] [review]
Precompile forms.js and preload UserAgenOverrides.jsm, v2

Review of attachment 710432 [details] [diff] [review]:
-----------------------------------------------------------------

::: modules/libpref/src/init/all.js
@@ +27,5 @@
>  // This pref exists only for testing purposes. In order to disable all
>  // overrides by default, don't initialize UserAgentOverrides.jsm.
>  pref("general.useragent.site_specific_overrides", true);
>  
> +// Enable UA overrides for shipping product code.

I know what the pref does, but I fail to understand the English sentence. This may be just me, however ;)

Comment 16

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Pushed with more verbose comment

https://hg.mozilla.org/integration/mozilla-inbound/rev/2716ae058067

Comment 17

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/2716ae058067

Comment 18

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Comment on attachment 710432 [details] [diff] [review]
Precompile forms.js and preload UserAgenOverrides.jsm, v2

Low-risk patch that significantly reduces time spent on critical startup path in app processes.

Comment 19

[Lukas Blakk [:lsblakk] use ?needinfo]

Comment on attachment 710432 [details] [diff] [review]
Precompile forms.js and preload UserAgenOverrides.jsm, v2

Please go ahead with uplift to the tip of mozilla-b2g18 branch for v1.0.1

Comment 20

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

https://hg.mozilla.org/releases/mozilla-b2g18/rev/e4ffa0660d88
https://hg.mozilla.org/releases/mozilla-b2g18/rev/d3401943481d