On 07/12/2013 11:54 AM, Jonathan Griffin wrote:
I think there's some confusion here. The tests that are reporting to
datazilla have nothing to do with TBPL or tryserver, and little to do with
buildbot, except that it uses buildbot nightly builds. These perf tests are
run using a completely separate system.
Are you trying to reproduce the same configuration that's used in the perf
tests?
I did once. My question is general for people who might be interested in
bisecting B2G. Can the try be used to do bisection of B2G? Or they have to
do the bisection by them self after being stuck between 2 mozilla-central
images?
Jonathan
On 7/12/2013 11:42 AM, Nicolas B. Pierron wrote:
On 07/12/2013 02:06 AM, Dave Hunt wrote:
On 7/11/13 6:35 PM, Nicolas B. Pierron wrote:
What is the configuration of the try server, does it pick the
corresponding gaia version from the mysterious gaia-changeset which is
in the Gecko's repository, or it just use the lastest commit of gaia?
I checks out the latest from Gaia's master branch and flashes that on top of
the nightly mozilla-central build. This allows us to detect Gaia regressions
as well as Gecko ones. The Gecko/Gaia revisions are visible in datazilla for
each result.
For the v1-train results, it does the same but flashes latest from Gaia's
v1-train on top of the latest mozilla-b2g18 build.
I am not sure to understand. If I push an old changeset of Gecko, I will
then have the latest Gaia?
The question is, can we use the try server to generate images which are
corresponding to what had already been built on inbound, knowing that we
do not keep inbound images on pvtbuild, but we do keep a few days of try's
images on pvtbuild.
--
Nicolas B. Pierron
_______________________________________________
dev-b2g mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-b2g
