Also tweak some things in the task definition, and rename the mozharness
config file to make mozharness happy.
--HG--
rename : testing/mozharness/configs/builds/releng_base_mac_cross_builds.py => testing/mozharness/configs/builds/releng_base_mac_64_cross_builds.py
extra : commitid : JCbwc6Wh9Oo
extra : rebase_source : 9b99fa75e225e8e2e10817155e481ba0493ee2fd
Messing up with shared cached object directories for different branches
often triggers build bustage. Make the object directory unique for each
branch.
Messing up with shared cached object directories for different branches
often triggers build bustage. Make the object directory unique for each
branch.
--HG--
extra : rebase_source : cf9b2f774133b17731c0bd125eccd1dbe8295d4e
When setting up the Firefox UI tests repository, it generates two binaries, one of them
called firefox-ui-update. Calling this binary on Windows triggers the UAC prompt for the
Release Engineering Windows test machines.
If instead we use Python to call a script that calls the same functionality we don't
get this UAC prompt.
We also move around some pre action checks to the right place. This allows us to not need to
specify --firefox-ui-branch when we're not running the checkout action.
--HG--
extra : commitid : HLO0z0sX6XV
extra : rebase_source : fa442b572b5b2570c7f830ef0d53f01474000853
extra : histedit_source : 567d2a79eab38593b658b17757ba689fdfcb8c8e
This synchronise the TALOSDATA regex with the one used in perfherder.
Also this ensure that we only see that message once.
--HG--
extra : commitid : JvuOhAtBUhJ
extra : rebase_source : bb733a6a5c74e44eb3fc2720964f1f0dbf220e0f
The config files under testing/config/mozharness were created so that certain mozharness options
such as test harness arguments could ride the trees, simplifying a lot of logic in mozharness.
But now that mozharness itself is in-tree, these configs no longer serve any purpose. Instead
they are merged into the main configs at testing/mozharness/configs.
--HG--
extra : commitid : KT4Ulv5TsiS
extra : rebase_source : 3cd1284680ddbddf0f1712a985a07579b3905849
extra : source : 28fa968ea46ca3e232b8cfbb4a88e9a492e71b16
This required moving the custom mozharness configs out of the builder image
(which puts them in ~, where mozharness no longer looks) and moving them into
mozharness itself.
--HG--
rename : testing/docker/desktop-build/configs/disable_signing.py => testing/mozharness/configs/disable_signing.py
rename : testing/docker/desktop-build/configs/platform_supports_post_upload_to_latest.py => testing/mozharness/configs/platform_supports_post_upload_to_latest.py
extra : commitid : 8FSj0uwFfbE
extra : rebase_source : 949bb94ca58324444981b81e9e9640ab00d63196
extra : amend_source : 40152e06e587728500800de02a437389038fe6ea
extra : source : e3c15b9f89b3c331805b4f1a83f62805f32bbb9a
The config files under testing/config/mozharness were created so that certain mozharness options
such as test harness arguments could ride the trees, simplifying a lot of logic in mozharness.
But now that mozharness itself is in-tree, these configs no longer serve any purpose. Instead
they are merged into the main configs at testing/mozharness/configs.
--HG--
extra : commitid : KJuwMFZEsNB
extra : rebase_source : ebad3073e7ea0e9575582b3355ae17909f7f39a7
* We move the generic platform specific config files under releng_infra_configs/ and remove
the word 'generic' from their file name
* Added --limit-locales to firefox_ui_updates.py and moved the dry_run break
* We improved flake8 score on firefox_ui_updates.py
* Small fix for vcstools.py
--HG--
rename : testing/mozharness/configs/generic_releng_linux.py => testing/mozharness/configs/releng_infra_configs/linux.py
rename : testing/mozharness/configs/generic_releng_linux64.py => testing/mozharness/configs/releng_infra_configs/linux64.py
rename : testing/mozharness/configs/generic_releng_macosx64.py => testing/mozharness/configs/releng_infra_configs/macosx64.py
rename : testing/mozharness/configs/generic_releng_win32.py => testing/mozharness/configs/releng_infra_configs/win32.py
rename : testing/mozharness/configs/generic_releng_win64.py => testing/mozharness/configs/releng_infra_configs/win64.py
extra : commitid : LQUsjvWCwAu
When we ./build.sh blobfree, we have a new zip file called
<target>.blobfree-dist.zip which contains a blobfree image. We copy this
file to public artifact.
We also refactor the build scripts a little bit to deduplicate the post
build code.
--HG--
extra : commitid : JU2n5YDGBfa
If we ./build.sh blobfree, move the zip file to public upload
dir.
Also, we don't need a blobfree specific config file anymore.
--HG--
extra : commitid : Gp8QmmLVcju
This also remove the useless xperf argument given for xperf tests
--HG--
extra : commitid : 8wJeWrMhtiA
extra : rebase_source : cf66053eeaefb8eed855b792fdcd7b35c12bc096
This prevents mach try's filtering of the master mochitest manifest from
interacting with the plugins subsuite by selecting by manifest rather than
test path. The way we run mochitest other could use a lot of cleanup --
this is a quick fix to prevent misleading output on treeherder when using
mach try.
--HG--
extra : commitid : AqQfTnQKIYv
When a local fails on Firefox UI tests we output the mozharness
command which a person could run on their local development machine
to reproduce the exact failure that we encoutered.
As parameters and options changes over time, we can fall out of date.
This patchs tries to reduce how we can fall out of date by
constructing the command rather than having a hardcoded output.
Another aspect which this patch solves is that for Firefox releases
we have to do some string manipulation to determine where we can
download Firefox crash symbols.
We've modified the script to output in the log the candidate URLs
that we're evaluating. This is useful if anyone wants to determine
why --symbols-path was or was not included. For instance, older candidate
directories on FTP are deleted after some time, hence, deleting the
crash symbols available.
Another matter we fix is that we include in the output what command
was run internally to the Firefox UI test binaries.
This is useful if anyone wants to determine that the right command
are arguments were used.
--HG--
extra : histedit_source : 2036448decbbe40fe18a018ff05f72928207caaf