663894 - Tracking bug for build and release of Firefox 5.0

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[[e-mail] (:bhearsum)]

Waiting on a go

Comment 1

[[e-mail] (:bhearsum)]

Attachment: stop caring about isFinalRelease in tagging

Comment 2

[[e-mail] (:bhearsum)]

Comment on attachment 538955 [details] [diff] [review]
stop caring about isFinalRelease in tagging

Landed this.

Comment 4

[[e-mail] (:bhearsum)]

Attachment: mozilla-release release config

I decided not to set nextVersion to 5.0.1, because it's unlikely we'll be having one, and thought it best to avoid confusion.

I enabled partner repacks here, and disabled them in 2.0.

I turned l10nMerge off, but I haven't got confirmation from anyone that we want to do that, yet.

Similar for l10n-changesets, I'm not sure if that's going to be our final set of locales.

Comment 5

[[e-mail] (:bhearsum)]

Attachment: basic mozilla-release patcher config

I used this patcher config in my staging run, and it worked as expected: The bumper changed the URLs to staging ones, added the <5.0> section, re-arranged a couple of things, and added a few "beta" things (beta-dir, beta-url). Because I don't have "beta" listed as a channel, these aren't actually used, but they're harmless. We should fix patcher-config-bump.pl to support this scenario better at some point.

One other caveat: the updates builder will fail, being unable to push the aus2.beta snippets. This will be after it has uploaded and pushed the test snippets, and after it has uploaded the release snippets. I'm not really sure how to fix this issue in time for the release, so I'm planning to just live with it, and manually kick off the post-updates tasks, and manually run the snippet comparison step.

I'm still waiting on some of the update verify chunks to finish, but the logs I've looked at have looked fine.

Comment 6

[[e-mail] (:bhearsum)]

Attachment: add update verify configs

Comment 7

[[e-mail] (:bhearsum)]

Comment on attachment 539279 [details] [diff] [review]
basic mozilla-release patcher config

Turns out we want to go to beta with this release, so this needs a minor tweak.

Comment 8

[[e-mail] (:bhearsum)]

Attachment: patcher config with beta channel enabled

Comment 9

[[e-mail] (:bhearsum)]

Comment on attachment 539344 [details] [diff] [review]
patcher config with beta channel enabled

Sorry for the churn, after discussion on IRC, it seems like the best thing to do is let the automation create the 4.0rc1/rc2/4.0.1 -> 5.0 snippets on release + test channels, and do the 5.0b* -> 5.0 ones by hand, in a separate patcher config (probably mozBeta). So, my previous patch ends up being correct for mozRelease.

Comment 10

[[e-mail] (:bhearsum)]

Attachment: mozilla-release release configs, v2

Same as the last version, except enable "mozilla-release" on pm01 instead of "mozilla-beta".

Comment 11

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

Comment on attachment 539325 [details] [diff] [review]
add update verify configs

>diff --git a/release/updates/mozRelease-firefox-linux.cfg b/release/updates/mozRelease-firefox-linux.cfg
...
>+# 4.0rc1 linux - full check
>+release="4.0" product="Firefox" platform="Linux_x86-gcc3" build_id="20110303140816" locales="de en-US ru" channel="betatest" patch_types="complete" from="/firefox/releases/4.0/linux-i686/%locale%/firefox-4.0.tar.bz2"

Looks like the wrong path in from, should be .../releases/4.0rc1/.... instead of /4.0/ - same issue in all four configs. Looks like that problem is in moz20-firefox-<platform>.cfg too, and moz20-firefox-linux.cfg says 'linux64 in comments'. r+ if you fix them all up.

Comment 12

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

Comment on attachment 539365 [details] [diff] [review]
mozilla-release release configs, v2

Review of attachment 539365 [details] [diff] [review]:
-----------------------------------------------------------------

Looks fine to me. We should switch this
  releaseConfig['ausServerUrl']        = 'https://aus2.mozilla.org'
to aus3.m.o for 2.0 and beta since that's where the apps are querying. It's possible that will break https if the cert issuer is unknown on our build slaves, in which case we can revert to aus2 as a short term fix. Bug 658066 is going to redirect aus2 to aus3 anyway.

::: mozilla/release-firefox-mozilla-2.0.py
@@ +61,3 @@
>  }
>  
>  # Platform configuration

Did you mean to toggle
  releaseConfig['doPartnerRepacks']    = True
to False here too?

Comment 13

[Rail Aliiev [:rail]]

Comment on attachment 539365 [details] [diff] [review]
mozilla-release release configs, v2

* partner repacks should be disabled in *-2.0.py (releaseConfig['doPartnerRepacks'] = False)
* the patch may be bitrotten, 5.0b7 is the last version

Comment 14

[Rail Aliiev [:rail]]

Comment on attachment 539325 [details] [diff] [review]
add update verify configs

I'll double Nick here. Just make sure that 4.0rc1 has from="...4.0rc1..."

Comment 15

[[e-mail] (:bhearsum)]

(In reply to comment #11)
> Comment on attachment 539325 [details] [diff] [review] [review]
> add update verify configs
> 
> >diff --git a/release/updates/mozRelease-firefox-linux.cfg b/release/updates/mozRelease-firefox-linux.cfg
> ...
> >+# 4.0rc1 linux - full check
> >+release="4.0" product="Firefox" platform="Linux_x86-gcc3" build_id="20110303140816" locales="de en-US ru" channel="betatest" patch_types="complete" from="/firefox/releases/4.0/linux-i686/%locale%/firefox-4.0.tar.bz2"
> 
> Looks like the wrong path in from, should be .../releases/4.0rc1/....
> instead of /4.0/ - same issue in all four configs. Looks like that problem
> is in moz20-firefox-<platform>.cfg too, and moz20-firefox-linux.cfg says
> 'linux64 in comments'. r+ if you fix them all up.

Good catch, will do.

(In reply to comment #12)
> Comment on attachment 539365 [details] [diff] [review] [review]
> mozilla-release release configs, v2
> 
> Review of attachment 539365 [details] [diff] [review] [review]:
> -----------------------------------------------------------------
> 
> Looks fine to me. We should switch this
>   releaseConfig['ausServerUrl']        = 'https://aus2.mozilla.org'
> to aus3.m.o for 2.0 and beta since that's where the apps are querying. It's
> possible that will break https if the cert issuer is unknown on our build
> slaves, in which case we can revert to aus2 as a short term fix. Bug 658066
> is going to redirect aus2 to aus3 anyway.

Sounds good, will do.

> ::: mozilla/release-firefox-mozilla-2.0.py
> @@ +61,3 @@
> >  }
> >  
> >  # Platform configuration
> 
> Did you mean to toggle
>   releaseConfig['doPartnerRepacks']    = True
> to False here too?

Yeah, whoops.


Thanks for the great reviews guys!

Comment 16

[[e-mail] (:bhearsum)]

Attachment: updated update verify configs

Fixed all of the 4.0 vs 4.0rc1 stuff, and the comments in the 2.0 linux config.

Comment 17

[[e-mail] (:bhearsum)]

Attachment: updated release config

Fixed AUS URL and 2.0 partner repack config.

Comment 18

[Staś Małolepszy :stas]

Re l10n: please use the same changesets as for Firefox 5 Beta Build 6 and 7.

Comment 19

[[e-mail] (:bhearsum)]

Thank you Staś, will do!

Comment 20

[[e-mail] (:bhearsum)]

Kev tells me that we need to provide complete updates only for Yahoo! partner builds. We'll need to manually generate snippets for the following partner channels (all platforms):
release-cck-yahoode
release-cck-yahoode
release-cck-yahoode
release-cck-yahoodek
release-cck-yahoode
release-cck-yahoode
release-cck-yahooau
release-cck-yahooauk
release-cck-yahooca
release-cck-yahoocak
release-cck-yahoogb
release-cck-yahoogb
release-cck-yahoogb
release-cck-yahoogbk
release-cck-yahooin
release-cck-yahooink
release-cck-yahoomy
release-cck-yahoomyk
release-cck-yahoonz
release-cck-yahoonzk
release-cck-yahooph
release-cck-yahoophk
release-cck-yahoosg
release-cck-yahoosgk
release-cck-yahoo
release-cck-yahoo
release-cck-yahook
release-cck-yahooxe
release-cck-yahooxe
release-cck-yahooar
release-cck-yahooark
release-cck-yahooes
release-cck-yahooes
release-cck-yahooesk
release-cck-yahoomx
release-cck-yahoomxk
release-cck-yahooe1
release-cck-yahooe1k
release-cck-yahoofr
release-cck-yahoocf
release-cck-yahoocfk
release-cck-yahoofr
release-cck-yahoofr
release-cck-yahoofrk
release-cck-yahooid
release-cck-yahooidk
release-cck-yahooit
release-cck-yahooit
release-cck-yahooitk
release-cck-yahookr
release-cck-yahookrk
release-cck-yahoobr
release-cck-yahoobrk
release-cck-yahooth
release-cck-yahoothk
release-cck-yahoovn
release-cck-yahoovnk
release-cck-yahoohk
release-cck-yahoohkk
release-cck-yahootw
release-cck-yahootwk

We did this once before, https://bugzilla.mozilla.org/show_bug.cgi?id=554992 has some notes on it.

Comment 21

[[e-mail] (:bhearsum)]

Attachment: update mozBeta patcher config for 5.0b* -> 5.0

I decided not to do partials for 5.0b7 -> 5.0, because we don't have many users there, and I don't know what the implications of having two sets of partials will be.

Comment 22

[[e-mail] (:bhearsum)]

Attachment: 5.0b* -> 5.0 update verify configs

I bumped these manually with update-verify-bump.pl, and then dropped "partial" from the list of patch types, because we're not providing them.

Comment 23

[[e-mail] (:bhearsum)]

Attachment: enable xulrunner

Comment 24

[[e-mail] (:bhearsum)]

Order of operations for pushing to mirrors:
- Push files to internal mirrors only
- Push index.html files to full set of mirrors
- <wait awhile so they all get them>
- Push files to full set of mirrors

We can probably do the first two points around the same time. We'll need to add something like this to the rsync exclude file to enable us to push index files without pushing the builds themselves.
+ firefox
+ firefox/releases
+ firefox/releases/5.0
+ firefox/releases/5.0/*/
+ firefox/releases/5.0/*/*/
+ firefox/releases/5.0/*/*/*/
+ firefox/releases/5.0/**index.html
- firefox/releases/5.0/**

Comment 25

[[e-mail] (:bhearsum)]

Attachment: add 4.0 betas to mozBeta patcher config

Anthony reminded me that we need to provide these beta users with updates, too. Contrary to earlier Betas, I'm using the mozBeta patcher config, because we'll want to continue bringing these users forward to the latest beta when we hit 6.0b1.

Comment 26

[[e-mail] (:bhearsum)]

Attachment: add 4.0b8+ to update verify configs

I didn't add 4.0rc1, 4.0rc2, or 4.0.1 here, because they're already getting tested in mozRelease-*, no need to duplicate that effort.

Comment 27

[[e-mail] (:bhearsum)]

Attachment: proposed index.html to push to mirrors

Christian, I'd like to push index files to the mirrors today, to make sure they all pick it up before we push the rest of the files on Monday. This is basically the same thing we used for Firefox 4 with updated verbiage and links.

Comment 28

[[e-mail] (:bhearsum)]

When we ship on Tuesday, we'll be throttling updates to 33% (that is to say, 33% of background update checks will receive an update, the other 67% will not).

Comment 29

[LegNeato]

Comment on attachment 540139 [details] [diff] [review]
proposed index.html to push to mirrors

Looks good, I tried it locally and it looked right

Comment 30

[[e-mail] (:bhearsum)]

Attachment: bump 3.6 branch throttling

AUS config patch to bump 3.6 branch throttling. To be landed on release day.

Comment 31

[[e-mail] (:bhearsum)]

Attachment: update AUS throttling

Per e-mail, 3.6.18 -> 4.0.1 is to be unthrottled, so we bump the throttling there to the next version. 4.0, 4.0.1 -> 5.0 is to be throttled at 33% to start with.

Comment 32

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

Comment on attachment 540660 [details] [diff] [review]
update AUS throttling

lgtm

Comment 33

[[e-mail] (:bhearsum)]

Attachment: update 5.0 throttling patch to throttle 4.0 and 4.0.1

Comment 34

[[e-mail] (:bhearsum)]

Comment on attachment 540679 [details] [diff] [review]
update 5.0 throttling patch to throttle 4.0 and 4.0.1

Landed this. Needs to get updated close to release time, before 4.0 -> 5.0 snippets are pushed.

Comment 35

[[e-mail] (:bhearsum)]

Attachment: patcher config for 3.5.x -> 5.0 updates

Per e-mail, we'll be updating most of the 3.5 releases to 5.0 with unadvertised manual updates. I decided not to do the alphas, betas, or RCs because I didn't think it was worth the extra snippets.

In order to block PPC users of releases earlier than 3.5.16 we'll need to add a section for 3.5 here: http://mxr.mozilla.org/mozilla/source/webtools/aus/xml/index.php#96.

Comment 36

[[e-mail] (:bhearsum)]

Attachment: 3.6.x -> 5.0 patcher config

Like 3.5, we're updating 3.6 users to 5.0 with an unadvertised manual update. Unlike 3.5, we can't block PPC users through index.php, because we want to be able to update them to 3.6.19 at a later time. Instead, we won't be updating any Mac builds that we can't identify by Build Target, which is everything prior to 3.6.12.

Comment 37

[[e-mail] (:bhearsum)]

Attachment: update verify config updates for new MUs

Comment 38

[[e-mail] (:bhearsum)]

Attachment: add 3.5 and 3.6 releases to AUS configs

Comment 39

[Chris AtLee [:catlee]]

Comment on attachment 540827 [details] [diff] [review]
add 3.5 and 3.6 releases to AUS configs

get rid of the extra 3.6.19 line and you're golden

Comment 40

[[e-mail] (:bhearsum)]

Attachment: block PPC users on 3.5

This isn't 100% ideal, but 3.5 users can't make it past 3.6 anyways...so I don't see great harm in doing it...the benefit of moving x86 users to 5.0 outweighs any harm IMHO.

Comment 41

[[e-mail] (:bhearsum)]

Attachment: drop as, ka, mn, oc

I copied the update verify lines from the minor update verify configs, and as such, got some locales in them that we don't have in 5.0. All of: as, ka, mn, and oc have been dropped since 3.5/3.6, so we shouldn't test for them here...

Comment 42

[[e-mail] (:bhearsum)]

Attachment: throttling patch w/out duplicate 3.6.19 entry

Comment 43

[[e-mail] (:bhearsum)]

Comment on attachment 540865 [details] [diff] [review]
block PPC users on 3.5

Via IRC:
[12:42] <rhelmer> bhearsum: yeah that patch lgtm, sorry for delay had to step away for a sec

Comment 44

[[e-mail] (:bhearsum)]

Attachment: update AUS throttling v3, remove 4.0* throttles

Per r-d, we should be unthrottling 4.0 and 4.0.1 today. May as well batch it together with the 3.5/3.6 stuff.

Comment 45

[[e-mail] (:bhearsum)]

Comment on attachment 540865 [details] [diff] [review]
block PPC users on 3.5

We decided not to ship these updates last minute, so this patch isn't going to be used :(.

Comment 46

[[e-mail] (:bhearsum)]

Comment on attachment 541417 [details] [diff] [review]
update AUS throttling v3, remove 4.0* throttles

I landed this patch, except for the 3.5 parts, because we decided not to ship that update.

Comment 47

[[e-mail] (:bhearsum)]

Attachment: unthrottle 3.6.2 - 3.6.17

Per release-drivers, 3.6.2 -> 3.6.17 should go to 3.6.18 instead of 5.0 directly, so we can theoretically move people to a more secure version sooner. This requires some restoring of snippets from backup, and then landing this patch.

Comment 48

[[e-mail] (:bhearsum)]

We finished up last week. See also: bug 666953 for a last minute XULRunner fix.