455089 - No OSX10.4 unittest boxes running on Firefox tinderbox

Status: RESOLVED WONTFIX

(Release Engineering :: General, defect, P3)

Description

[Boris Zbarsky [:bzbarsky]]

All the Mac test boxes on the Firefox tinderbox are running Leopard.  It would be great to have a Tiger box, so that we will be able to catch regressions that are Tiger-specific.  This includes regressions in handling of fonts (which works a little differently due to Tiger bugs that we work around), and gfx regressions.

Comment 1

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

triage to future for now.

Comment 2

[Boris Zbarsky [:bzbarsky]]

For what it's worth, this would be of most use _before_ the 3.1 release, since the goal is to catch regressions before we ship them...

Comment 3

[Samuel Sidler (old account; do not CC)]

Yeah, what's the chance we can get one setup in the next month? I know RelEng is pretty busy, but it seems like this is important (even if we only get one box instead of the regular dozen or so).

Comment 4

[Samuel Sidler (old account; do not CC)]

We likely should have machines that are PPC as well. (see bug 457486)

Comment 5

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

See also bug 447999/bug 456606 comment 0.

Comment 6

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

We have two 10.4 boxes on Firefox3.0, and two 10.5 boxes on Firefox - all Intel machines. Since PPC trees don't grow on trees these days, perhaps the best way forward is have one of OS on each tree. 

Lukas, is that a relatively simple change to the master configs ?

Comment 7

[Samuel Sidler (old account; do not CC)]

This bug was specifically about getting 10.4 boxes on the Firefox tree given there are differences in tests that run on each OS. PPC was additional and separate (I should probably file a bug for it).

Comment 8

[Samuel Sidler (old account; do not CC)]

Oh, and in case it wasn't clear, 10.4 runs on Intel machines.

Comment 9

[Lukas Blakk [:lsblakk] use ?needinfo]

Attachment: Swap out a 10.4 slave from 1.9 Production, replace with 10.5

Comment 10

[Lukas Blakk [:lsblakk] use ?needinfo]

Attachment: Swap out a 10.5 slave from M-C Unittest, replace with 10.4

Comment 11

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

Comment on attachment 346767 [details] [diff] [review]
Swap out a 10.5 slave from M-C Unittest, replace with 10.4

These patches look fine to me but I just thought of a problem. We should probably upgrade/downgrade XCode to match the other machines on their new branch. That'll probably go OK on bm-xserve21 moving to Firefox, since you can run XCode 3.1 on Tiger; but I betcha XCode 2.2.1 doesn't install on Leopard. IIRC 3.0 is the minimum there.

Comment 12

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

(In reply to comment #11)
> (From update of attachment 346767 [details] [diff] [review])
> These patches look fine to me but I just thought of a problem. We should
> probably upgrade/downgrade XCode to match the other machines on their new
> branch. That'll probably go OK on bm-xserve21 moving to Firefox, since you can
> run XCode 3.1 on Tiger; but I betcha XCode 2.2.1 doesn't install on Leopard.
> IIRC 3.0 is the minimum there.

You've got that mostly backwards; Xcode 2.n<5 will run on 10.4, Xcode 2.5 will run on 10.4 and 10.5, but Xcode 3.x will only run on 10.5.

If you do want to use Xcode 2.5 on 10.5, you'll need to uninstall Xcode 3, because the command-line toolchains default to the Xcode 3 toolchain regardless (you only get the 2.x toolchain in the Xcode 2.5 app UI).

Comment 13

[Lukas Blakk [:lsblakk] use ?needinfo]

So I have a 10.4 Intel mini doing mozilla-central builds and running unittests - 

You can watch it on the MozillaTest Tinderbox page, the machine is called:

MacOSX Darwin 8.8.4 bm-stage-osx-01 dep unit test

It will stay there for now, but eventually move into production.

Comment 14

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

bz: I've cc'd you on related bug#471006.

Comment 15

[Lukas Blakk [:lsblakk] use ?needinfo]

The standalone unittests are now closed, which means that this buildslave is idle and not running unittests.

For now, we'll keep this machine so that we can run unittests against downloaded builds on it once bug 421611 is wrapped up.

Comment 16

[Markus Stange [:mstange]]

What's the status here? I'm wondering whether it's worth to write a test for bug 507845, which is 10.4-only.

Comment 17

[Boris Zbarsky [:bzbarsky]]

Yes; there are non-Firefox tinderboxes running 10.4.

Comment 18

[Mark Banner (:standard8)]

(In reply to comment #17)
> Yes; there are non-Firefox tinderboxes running 10.4.

I'm not sure which boxes you are thinking of. If you're thinking Thunderbird then we dropped all our 10.4 boxes recently. If you're thinking SeaMonkey then the 10.4 unit test box is only running on 1.9.1 branch and I think I heard KaiRo say he was thinking of dropping that one.

Comment 19

[Boris Zbarsky [:bzbarsky]]

I was thinking Seamonkey.

If we plan to keep supporting 10.4, then we should be writing tests for things that fail there, and we should be testing it (what this bug is about).  Period.

Since we plan to support it for at least another year at the very least (lifetime of the 1.9.2 branch)...

Comment 20

[Robert Kaiser]

SeaMonkey will drop the 10.4 unit test box as soon as the 10.5 VMs are running stable (Parallels has proven to be somewhat problematic in OSX VM stability).

We already have a 10.4 Xserve (cb-xserve02) we used for SeaMonkey builds and nightlies for a long time, which is not in use any more, we had in mind to give it back to IT for taking down or to whatever use comes up, it probably wouldn't be too hard to repurpose it as a 10.4 unit test slave for Firefox if wanted.

Comment 21

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

(In reply to comment #20)
> SeaMonkey will drop the 10.4 unit test box as soon as the 10.5 VMs are running
> stable (Parallels has proven to be somewhat problematic in OSX VM stability).
> 
> We already have a 10.4 Xserve (cb-xserve02) we used for SeaMonkey builds and
> nightlies for a long time, which is not in use any more, we had in mind to give
> it back to IT for taking down or to whatever use comes up, it probably wouldn't
> be too hard to repurpose it as a 10.4 unit test slave for Firefox if wanted.

Kairo, if this machine has become available, we'll happily use it for our geriatric-master work (bug#521888). Do you happen to know if cb-xserve02 is Intel or PPC?


(Pushing bug over to jhford, as he's working on other bugs in that area already)

Comment 23

[John Ford [:jhford] CET/CEST Berlin Time]

mozilla-central does have 10.4 compatible builds any more.   530020 tracks the requirements for us to be able to run unit tests for mozilla-1.9.2 on 10.4.

Comment 24

[John Ford [:jhford] CET/CEST Berlin Time]

moving to future until dependent bugs are resolved

Comment 25

[Robert Kaiser]

(In reply to comment #21)
> Kairo, if this machine has become available, we'll happily use it for our
> geriatric-master work (bug#521888). Do you happen to know if cb-xserve02 is
> Intel or PPC?

joduinn/jford:
See bug 492224 - it's PPC from all I know and it is free for grabbing, as I already told you there :)

Comment 26

[Chris Cooper [:coop] (he/him)]

Mass move of bugs from Release Engineering:Future -> Release Engineering. See
http://coop.deadsquid.com/2010/02/kiss-the-future-goodbye/ for more details.

Comment 27

[Lukas Blakk [:lsblakk] use ?needinfo]

Resolving WONTFIX since we don't support 10.4 any longer.

Comment 28

[Samuel Sidler (:ss)]

(If we just wait a year and a half, bugs might be irrelevant. Yay, progress!)