Closed Bug 813613 Opened 12 years ago Closed 11 years ago

Build Firefox 17.0.5ESR using mock

Categories

(Release Engineering :: Release Requests, defect)

x86
Linux
defect
Not set
normal

Tracking

(firefox-esr17+ fixed)

RESOLVED WONTFIX
Tracking Status
firefox-esr17 + fixed

People

(Reporter: jhopkins, Assigned: bhearsum)

References

Details

(Keywords: verifyme)

Attachments

(3 files, 2 obsolete files)

      No description provided.
Blocks: 805587
I'm on the hook for 17.0.1 ESR.
Assignee: nobody → bhearsum
Blocks: 809319
No longer blocks: 805587
Blocks: 805587
still blocking for same ESR date, but different version now (17.0.2 ESR)
Blocks: 815763
No longer blocks: 809319
Summary: Build Firefox 17.0.1ESR using mock → Build Firefox 17.0.2ESR using mock
Is this 100% ready to go? We're about 48 hours from go to build.
Oops, of course it's not ready to go yet, there's no patch!
Attached patch use mock for mozilla-esr17 (obsolete) (deleted) — Splinter Review
This should be landed in advance of the go to build, because I want to make sure that dep/nightly builds are working fine before we build a release with mock.
Attachment #697053 - Flags: review?(catlee)
Attachment #697053 - Flags: review?(catlee) → review+
Attachment #697053 - Flags: checked-in+
in production, forcing some builds to verify it
The builds that I triggered went fine.
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → FIXED
Comment on attachment 697053 [details] [diff] [review]
use mock for mozilla-esr17

Backed out due to bug 826567.
Attachment #697053 - Flags: checked-in+ → checked-in-
We'll aim to have this for the next esr instead
Blocks: 825314
No longer blocks: 815763
Assignee: bhearsum → nobody
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Summary: Build Firefox 17.0.2ESR using mock → Build Firefox 17.0.3ESR using mock
Component: Release Engineering: Automation (Release Automation) → Release Engineering: Releases
Depends on: 827354
Assignee: nobody → coop
Moving this to 17.0.4 because it's too close to 17.0.3 for comfort.
Blocks: 837125
No longer blocks: 825314
Summary: Build Firefox 17.0.3ESR using mock → Build Firefox 17.0.4ESR using mock
I'm going to make sure this uplift happens right after we ship 17.0.3esr.
Assignee: coop → bhearsum
Attached patch fully enable mock for firefox esr17 (obsolete) (deleted) — Splinter Review
Attachment #697053 - Attachment is obsolete: true
Attachment #716537 - Flags: review?(jhopkins)
Comment on attachment 716537 [details] [diff] [review]
fully enable mock for firefox esr17

wrong version of gcc - need to fix this.
Attachment #716537 - Attachment is obsolete: true
Attachment #716537 - Flags: review?(jhopkins)
Attached patch use moz4 gcc (deleted) — Splinter Review
Need to test this in staging.
Attached patch bump in repo files (deleted) — Splinter Review
glandium gave me a way to verify the build. Things look good:
[root@dev-stage01 fesr2]# tar -jvxf /pub/mozilla.org/firefox/nightly/2013/02/2013-02-21-08-20-01-mozilla-esr17/firefox-17.0.3esrpre.en-US.linux-x86_64.tar.bz2
readelf -D -s *.so | grep UNIQUE
[root@dev-stage01 firefox]#
Attachment #716546 - Flags: review?(jhopkins)
Attachment #716557 - Flags: review?(jhopkins)
Attachment #716557 - Flags: review?(jhopkins) → review+
Attachment #716546 - Flags: review?(jhopkins) → review+
Comment on attachment 716557 [details] [diff] [review]
bump in repo files

Drivers, this patch will let us switch esr17 over to our new build slaves (larger pool, old slaves can be repurposed to help windows build wait times). It includes the new gcc built in bug 827354, so it does not cause a system requirements change.
Attachment #716557 - Flags: approval-mozilla-esr17?
Comment on attachment 716557 [details] [diff] [review]
bump in repo files

We discussed this at the channel meeting and agreed that we'd be comfortable landing/testing for the next ESR. Approving for the ESR branch.
Attachment #716557 - Flags: approval-mozilla-esr17? → approval-mozilla-esr17+
Attachment #716546 - Flags: checked-in+
Assigning this to Matt Wobensmith for QA. What should we keep an eye on with this change (ie. what are the high risk areas)?
Keywords: qawanted
QA Contact: bhearsum → mwobensmith
The only test that needs to be done (AFAIK) is to see whether the new builds still work on certain Linux versions. It looks like bug 823487 has some background on this. (I wasn't very involved with finding the fix, you may want to talk to glandium or jhopkins if that bug doesn't have the information required.)
Attachment #716557 - Flags: checked-in+
Attachment #717962 - Flags: review?(jhopkins)
Attachment #717962 - Flags: review?(jhopkins) → review+
Attachment #717962 - Flags: checked-in+
John Hopkins was working with somebody in https://bugzilla.mozilla.org/show_bug.cgi?id=832601 to test out his original builds with the new GCC, too. That person may be worth contacting if we can't test directly.

As far as I'm concerned, this is done.
Status: REOPENED → RESOLVED
Closed: 12 years ago12 years ago
Resolution: --- → FIXED
Ben - comment 20 mentions bug 823487, but that bug does not appear to be related to this one. Can you double check? I'm seeking info on what Linux versions should be targeted for testing. Thanks.
(In reply to Matt Wobensmith from comment #25)
> Ben - comment 20 mentions bug 823487, but that bug does not appear to be
> related to this one. Can you double check? I'm seeking info on what Linux
> versions should be targeted for testing. Thanks.

Whoops, apologies. That should be bug 827354.
Blocks: 837128
I've run passes on CentOS 6.3 and Debian 6, both 64bit. All seems OK.

Are there concerns for other versions of the above - e.g. Debian 5, and/or 32bit?

If not, we can call this bug verified.
(In reply to Matt Wobensmith from comment #27)
> I've run passes on CentOS 6.3 and Debian 6, both 64bit. All seems OK.
> 
> Are there concerns for other versions of the above - e.g. Debian 5, and/or
> 32bit?

Yes, it was specifically older versions of these distros that are a concern.
Comment on attachment 716546 [details] [diff] [review]
use moz4 gcc

This was backed out due to bustage in the chemspill 17.0.4esr release. Will be relanded once all the chemspills are done. Also need to land the mozilla-esr17 patch on the relbranches.
Attachment #716546 - Flags: checked-in+ → checked-in-
Attachment #717962 - Flags: checked-in+ → checked-in-
Comment on attachment 716557 [details] [diff] [review]
bump in repo files

Backed out in https://hg.mozilla.org/releases/mozilla-esr17/rev/0e3e0b0e735f since I blindly stumbled into the situation where the tree still thought it was building with moz4 and buildbot thought it would use moz3.
Attachment #716557 - Flags: checked-in+ → checked-in-
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Depends on: 849583
Comment on attachment 716546 [details] [diff] [review]
use moz4 gcc

Relanded now that chemspills are over. I'll be triggering some builds to verify once a reconfig puts these patches in production again.
Attachment #716546 - Flags: checked-in- → checked-in+
Attachment #717962 - Flags: checked-in- → checked-in+
Comment on attachment 716557 [details] [diff] [review]
bump in repo files

This has been relanded on default and the relbranches.
Attachment #716557 - Flags: checked-in- → checked-in+
Builds are looking good.
Status: REOPENED → RESOLVED
Closed: 12 years ago12 years ago
Resolution: --- → FIXED
Summary: Build Firefox 17.0.4ESR using mock → Build Firefox 17.0.5ESR using mock
(In reply to Ben Hearsum [:bhearsum] from comment #32)
> Comment on attachment 716557 [details] [diff] [review]
> bump in repo files
> 
> This has been relanded on default and the relbranches.

Some of us were talking yesterday and it was pointed out to me that we probably don't want to take a change like this in a chemspill from a risk/reward standpoint. I backed out on both relbranches because of this, just in case.
Blocks: 848749
As per conversation with Ben, our remaining concern w/r/t testing is that we want to verify that this build runs on a distro that has the following config:

GTK 2.10 to 2.17
Glib 2.12 to 2.21

This corresponds to the ESR system support page here:
https://www.mozilla.org/en-US/firefox/17.0.4/system-requirements/

Getting an older distro - that meets the above reqs - running in a virtualized environment (internally) has been difficult. Hence, we intend to ask someone who already has this config in place. Ben is reaching out to a community member who should have this config, and will update this bug shortly.
(In reply to Matt Wobensmith from comment #35)
> As per conversation with Ben, our remaining concern w/r/t testing is that we
> want to verify that this build runs on a distro that has the following
> config:
> 
> GTK 2.10 to 2.17
> Glib 2.12 to 2.21
> 
> This corresponds to the ESR system support page here:
> https://www.mozilla.org/en-US/firefox/17.0.4/system-requirements/
> 
> Getting an older distro - that meets the above reqs - running in a
> virtualized environment (internally) has been difficult. Hence, we intend to
> ask someone who already has this config in place. Ben is reaching out to a
> community member who should have this config, and will update this bug
> shortly.

I did this over here: https://bugzilla.mozilla.org/show_bug.cgi?id=832601#c9.
He responded, and the builds work fine:

(In reply to Camaleon from comment #10)
> (In reply to Ben Hearsum [:bhearsum] from comment #9)
> > Camaleon, if you don't mind, could you test one more build out for us?
> > https://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/latest-mozilla-esr17/
> > firefox-17.0.4esrpre.en-US.linux-x86_64.tar.bz2
> 
> Sure, but as I don't have a 64-bits system I tested this file, instead:
> 
> https://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/latest-mozilla-esr17/
> firefox-17.0.4esrpre.en-US.linux-i686.tar.bz2
> 
> By the way, it runs okay.
Status: RESOLVED → VERIFIED
Using mock, even with the new compiler, still caused runtime issues on some platforms:
> Alex Keybl wrote:
>> As we approach our next targeted ESR release date of 4/2/2013, we'd
>> like your help in smoke testing the latest pre-release builds. We're
>> not asking for a full qualification of the builds prior to release,
>> but rather minor exploratory testing of internal websites and
>> applications that our QA team wouldn't otherwise have access to.
>> Please directly email release-mgmt@mozilla.com or this list with any
>> critical regressions found. Note that this release will include an
>> upgrade of NSS from 3.13.6 to NSS 3.14.3 for user security. If you
>> utilize SSL client authentication with smartcards, please verify that
>> your users will not be impacted by this upgrade.
>> -Alex Keybl Release Manager
>> NOTE: Please do not deploy any of the builds linked to below - they
>> are pre-release software. Updates will not function correctly, and
>> these pre-release builds will not be supported by Mozilla.
>> ESR17 Windows:
>> https://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/latest-mozilla-esr17/firefox-17.0.4esrpre.en-US.win32.installer.exe
>> ESR17 Mac OS X:
>> https://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/latest-mozilla-esr17/firefox-17.0.4esrpre.en-US.mac.dmg
>> ESR17 Linux:
>> https://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/latest-mozilla-esr17/firefox-17.0.4esrpre.en-US.linux-i686.tar.bz2
> 
> This Linux FF17.0.5esr pre-release doesn't run on RHEL/CentOS 5 - the
> error I get is:
> 
> Couldn't load XRE functions.
> 
> Which means it can't load libxul.so :
> 
> libxpcom.so: /lib/libc.so.6: version `GLIBC_2.7' not found (required by
> ./libxul.so)
> 
> It also requires libgio-2.0.so.0 - which doesn't exist on CentOS 5
> 
> ESR 17.0.4 does run OK on CentOS 5


(Ignore the filenames - those builds were built this morning.) I'm backing this out *again*. I think it's about time we WONTFIX this bug, too.
Status: VERIFIED → REOPENED
Resolution: FIXED → ---
Backed out in https://hg.mozilla.org/build/buildbot-configs/rev/d42aa7b28de7 and https://hg.mozilla.org/build/buildbot-configs/rev/7b4d026f8a1d. Once the next reconfig happens I'll be backing out the mozilla-esr17 file, too.
Attachment #716546 - Flags: checked-in+ → checked-in-
Attachment #717962 - Flags: checked-in+ → checked-in-
Attachment #716557 - Flags: checked-in+ → checked-in-
I couldn't figure out what it meant that only Linux64 builds broke in https://tbpl.mozilla.org/?tree=Mozilla-Esr17&rev=187596cad342 (in a way that looks sort of like it means "the NSS update we got while we were building with mock does not build without mock"), so I closed esr17 and comm-esr17.
(In reply to Phil Ringnalda (:philor) from comment #40)
> I couldn't figure out what it meant that only Linux64 builds broke in
> https://tbpl.mozilla.org/?tree=Mozilla-Esr17&rev=187596cad342 (in a way that
> looks sort of like it means "the NSS update we got while we were building
> with mock does not build without mock"), so I closed esr17 and comm-esr17.

I filed this as bug 855263. Not sure what we're going to do yet.
(In reply to Ben Hearsum [:bhearsum] from comment #38)
> > 
> > This Linux FF17.0.5esr pre-release doesn't run on RHEL/CentOS 5 - the
> > error I get is:
> > 
> > Couldn't load XRE functions.
> > 
> > Which means it can't load libxul.so :
> > 
> > libxpcom.so: /lib/libc.so.6: version `GLIBC_2.7' not found (required by
> > ./libxul.so)
> > 
> > It also requires libgio-2.0.so.0 - which doesn't exist on CentOS 5
> > 
> > ESR 17.0.4 does run OK on CentOS 5

The latest 32 bit Linux FF17.0.5esr pre-release (dated 27th March) runs OK on CentOS 5
Blocks: 855263
Given all of the issues we've had with trying to do this, I don't think it's worthwhile investing any more time into. We'll just have to prop up the old build platform until esr17 dies.
Status: REOPENED → RESOLVED
Closed: 12 years ago11 years ago
Resolution: --- → WONTFIX
Product: mozilla.org → Release Engineering
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: