764684 - Deal with app.update.stage.enabled

Status: RESOLVED FIXED

(Firefox OS Graveyard :: General, defect, P1)

Description

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

Support for a different update path landed since the b2g updater was enabled.  I'm not 100% sure how this affected the updater protocol, but we need to figure out what to do with this new configuration option.

Comment 1

[Dietrich Ayala (:dietrich)]

Blocking for now - can you clarify what this means?

Comment 2

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

No, it was rumored this was affecting b2g updates, but I haven't confirmed nor do I know what this setting does.

Comment 3

[Dietrich Ayala (:dietrich)]

Who would know?

Comment 4

[Nick Thomas [:nthomas] (UTC+12)]

If you're still looking, Ehsan is your man. That pref came from bug 307181.

Comment 5

[JP Rosevear [:jpr]]

Ehsan, do you have some docs on this?

Comment 6

[(no longer active)]

Not really.  I thought that b2g doesn't use the updater, same way that Android doesn't.  My patches disabled background updates for b2g IIRC.  What is exactly broken here?

Comment 7

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

Define "updater"?  We use nsUpdateService to download updates and then the related infrastructure to verify and apply them to the filesystem ...

It was unclear whether this bug affected that, but Marshall might know by now.

Comment 8

[(no longer active)]

I'm talking about the updater application: http://mxr.mozilla.org/mozilla-central/source/toolkit/mozapps/update/updater/updater.cpp

If that is used on b2g, then we should discuss how it needs to handle the background updates mode.

Comment 9

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

I ported that to gonk (b2g).  We use it.

Comment 10

[(no longer active)]

(In reply to comment #9)
> I ported that to gonk (b2g).  We use it.

Oh, OK, I was not aware of that.  It would be useful for me and someone who knows how the updater works in b2g to have a quick meeting next week so that I can figure out what the b2g story looks like.

Two more questions: what is the exact failure?  And do all of the updater tests on mozilla-central pass with b2g (do we run those automatically somewhere?)?

Comment 11

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

(In reply to Ehsan Akhgari [:ehsan] from comment #10)
> (In reply to comment #9)
> > I ported that to gonk (b2g).  We use it.
> 
> Oh, OK, I was not aware of that.  It would be useful for me and someone who
> knows how the updater works in b2g to have a quick meeting next week so that
> I can figure out what the b2g story looks like.
> 

Do you mean the update checking, downloading, or applying?  All the core parts are 100% the same, but the UI hookup is different obviously.  http://mxr.mozilla.org/mozilla-central/source/b2g/components/UpdatePrompt.js fyi.

> Two more questions: what is the exact failure?  And do all of the updater
> tests on mozilla-central pass with b2g (do we run those automatically
> somewhere?)?

I haven't reproduced this, we heard through the grapevine that this was a problem.  No, we don't have those tests running yet.

Comment 12

[Marshall Culpepper [:marshall_law]]

FYI, It does look like this is blocking the updater from working in Gonk.

(In reply to Ehsan Akhgari [:ehsan] from comment #10)
> It would be useful for me and someone who
> knows how the updater works in b2g to have a quick meeting next week

I'm available most of the week, Let's start with IRC and we can move to Skype/Vidyo if needed


> Two more questions: what is the exact failure?  And do all of the updater
> tests on mozilla-central pass with b2g (do we run those automatically
> somewhere?)?

In my local testing, since gCanStageUpdates is false, 
nsIUpdateProcessor::processUpdates is never being called. We aren't currently explicitly setting app.update.stage.enabled to true (is there anything else required?)

FWIW, we are not currently running Mochitest or xpcshell tests, so I doubt any of the automated updater tests are running. jgriffin and mdas are working on getting those running, though.

Comment 13

[(no longer active)]

I talked to Marshall about this on IRC today.  In short, I don't think there's any big thing to fix for b2g.  We just need to make sure that /system is remounted RW before applying the update, and the stage directory lives on the same physical partition as the appdir.  Other than this I think most of the Linux code path should be usable here unmodified.

Comment 14

[Marshall Culpepper [:marshall_law]]

It looks like the only changes required for this bug will be setting app.update.stage.enabled to true, and not writing to update.test from the gCanStageUpdates getter. 

The remount changes are already being tracked in Bug 764683

Comment 15

[Marshall Culpepper [:marshall_law]]

Attachment: enable update staging - v1

Feedback: :cjones

Comment 16

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

Comment on attachment 648046 [details] [diff] [review]
enable update staging - v1

Sure.

Comment 17

[(no longer active)]

Comment on attachment 648046 [details] [diff] [review]
enable update staging - v1

Review of attachment 648046 [details] [diff] [review]:
-----------------------------------------------------------------

r=me assuming that the update process has been fully tested on b2g (with the remounting patches of course).

Comment 18

[Marshall Culpepper [:marshall_law]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/bbc3d49c19ba

Comment 19

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/bbc3d49c19ba