Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Mac OS X 10.7 testing

13 views
Skip to first unread message

John Ford

unread,
Oct 26, 2011, 10:40:00 AM10/26/11
to mozilla.dev.planning planning, dev-tree-...@lists.mozilla.org, dev-pl...@lists.mozilla.org, release
Hello!

We are getting close to having Mac OS X 10.7.2 slaves ready for production. These slaves will be running the same hardware spec as the new Rev4 10.6 minis. I have configured the slave with minimal modifications to defaults. We don't expect that all tests on 10.7 to be green on day one. I have verified that the machines run stable with a known good changeset of Mozilla Central.

The changeset that I am running on the machines is generating the following failures for opt builds:
-Mochitest 1/5 - 109477/7/877
-Mochitest 2/5 - 10608/2/346
-Mochitest 4/5 - 114421/2/113
-Mochitest Other - (chrome) 32840/17/141
-Reftest - 6880/5/264
-XPCShell - 1180/4

Fwiw, this means that 0.013% of tests fail on 10.7.

My plan is to finish getting the full compliment of 80 machines set up and running with the same coverage as our 10.6 slaves. We can hide the test suites that have constant orange, but I think we need to continue running these tests to get them green.

I am often asked if this change means that we are building on 10.7 or building with clang. This change is only for *testing* machines, which do not do builds. The builds we are testing are the same bits that we are testing on 10.6 test machines. There is no change (yet) to us building mozilla-central using gcc on 10.6.

My plan is to ask for these machines to be imaged later on today and to put them into production later this week/early next when we have master

Thanks,

John Ford

Ted Mielczarek

unread,
Oct 26, 2011, 11:03:14 AM10/26/11
to John Ford, mozilla.dev.planning planning, dev-pl...@lists.mozilla.org, release, dev-tree-...@lists.mozilla.org
On Wed, Oct 26, 2011 at 10:40 AM, John Ford <jhf...@mozilla.com> wrote:
> The changeset that I am running on the machines is generating the following failures for opt builds:
> -Mochitest 1/5 - 109477/7/877
> -Mochitest 2/5 - 10608/2/346
> -Mochitest 4/5 - 114421/2/113
> -Mochitest Other  - (chrome) 32840/17/141
> -Reftest - 6880/5/264
> -XPCShell - 1180/4
>
> Fwiw, this means that 0.013% of tests fail on 10.7.

If we're consistently failing the same tests, can you get us a list of
the failures? We should get bugs filed on these and either fix or skip
the tests.

-Ted

John Ford

unread,
Oct 26, 2011, 12:06:55 PM10/26/11
to Ted Mielczarek, mozilla.dev.planning planning, dev-pl...@lists.mozilla.org, release, dev-tree-...@lists.mozilla.org
Yes, I will file the bugs. Do you want one bug per test that fails or is it better to have one bug per test framework?

I'll also file a tracking bug to have all 10.7 tests green

John

Ted Mielczarek

unread,
Oct 26, 2011, 12:18:33 PM10/26/11
to John Ford, mozilla.dev.planning planning, dev-pl...@lists.mozilla.org, release, dev-tree-...@lists.mozilla.org
On Wed, Oct 26, 2011 at 12:06 PM, John Ford <jhf...@mozilla.com> wrote:
> Yes, I will file the bugs.  Do you want one bug per test that fails or is it better to have one bug per test framework?

"It depends". It might help to just get a list of failures per test
suite up somewhere, and get some developer feedback. It's possible we
have classes of failures where we can have a single bug per set of
tests. If you get this list up, I can help get developer attention on
it.

-Ted\
0 new messages