New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 714336 link

Starred by 2 users

Issue metadata

Status: Fixed
Owner:
Closed: Apr 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 0
Type: Bug

Blocked on:
issue 582094

Blocking:
issue 714976



Sign in to add a comment

GN analyze not detecing that browser_tests needs to run

Project Member Reported by jam@chromium.org, Apr 22 2017

Issue description

See this patch https://codereview.chromium.org/2832153003/#ps80001

and this try run (others also have this bug)
https://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_rel_ng/builds/436460

For some reason it didn't run browser_tests, even though it is affected.

The analyze output is https://luci-logdog.appspot.com/v/?s=chromium%2Fbb%2Ftryserver.chromium.mac%2Fmac_chromium_rel_ng%2F436460%2F%2B%2Frecipes%2Fsteps%2Fanalyze%2F0%2Fstdout which does have

"//chrome/test:browser_tests",
in the "test_targets" section but not in the
analyze output's test_targets
 
Weird. I don't get that output locally, analyze appears to run correctly.

I wonder if something has gotten messed up on the builders somehow?
Do you know of other CLs (different patches) that might have problems? I'll go looking for some. It looks like the master just restarted, so there's a bunch of builds that failed for unrelated reasons ...
nevermind, I have reproduced it, and it does seem to be happening on other CLs as well, e.g.: https://luci-logdog.appspot.com/v/?s=chromium%2Fbb%2Ftryserver.chromium.mac%2Fmac_chromium_rel_ng%2F436493%2F%2B%2Frecipes%2Fsteps%2Fanalyze%2F0%2Fstdout
Cc: kbr@chromium.org thakis@chromium.org
Status: Started (was: Assigned)
There seems to be some very weird interaction going on.

In https://codereview.chromium.org/2569333002 , thakis@ renamed 'tab_capture_end2end_tests_run' to 'tab_capture_end2end_tests', and that appears to break things.

'tab_capture_end2end_tests' is also defined in //testing/buildbot/gn_isolate_map.pyl .

The target is defined as a group in //chrome/test:BUILD.gn that depends on browser_tests, *and also* defined as an alias for browser_tests in gn_isolate_map.pyl, which means that the two labels conflict. I think something is getting confused and pruning things out. 


FWIW, tab_capture_end2end_tests isn't being run either. If you delete the _run target (which I think is correct), browser_tests gets picked up but tab_capture_end2end_tests doesn't, which may be a bug in GN. I need to figure out / remember how these two systems are supposed to interact :(.
actually it looks like tab_capture_end2end_tests is run in some situations:

https://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_rel_ng/builds/436537

I'm not sure which solutions produce which results yet; for all I know, it might be nondeterministic depending on the ordering of dict entries in python. I'm going to need to stare at this for a while to figure out what the right thing to do.

However, it's probably bad that browser_tests isn't being run, so I think the thing to do in the meantime is land a change that does run browser_tests, and temporarily disables the tab_capture tests to avoid the collision.
Cc: senorblanco@chromium.org enne@chromium.org
CL posted to temporarily (hopefully) disable tab_capture_end2end_tests and restore browser_tests: https://codereview.chromium.org/2828143007

+ the current gpu sheriffs ...
I've posted https://codereview.chromium.org/2838473002/ which I think is actually a correct fix, so hopefully the CL in #c7 won't be needed.
Project Member

Comment 9 by bugdroid1@chromium.org, Apr 23 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/75636899ba7c25c8b3b632989e8358987a3b0b50

commit 75636899ba7c25c8b3b632989e8358987a3b0b50
Author: dpranke <dpranke@chromium.org>
Date: Sun Apr 23 16:42:25 2017

Rework tab_capture_end2end_tests' test definition.

tab_capture_end2end_tests was previously defined as a target in
GN that depended on browser_tests, and as a target in the gn_isolate_map
as a "gpu_browser_test". In addition, the //testing/buildbot/ json
file generated by //content/test/gpu/generate_buildbot_json ended up
duplicating some of the logic that was in mb.py for handling
gpu_browser_tests.

The fact that the logic was scattered over multiple places ended up
confusing `mb analyze`, with the result that if you were trying to
run both browser_tests and tab_capture_end2end_tests, you might only
get one of them.

We need to reduce or eliminate the multiple different ways we have
of configuring test steps. Eventually we should push as much of this
logic as we can into the BUILD.gn files, I think, but for now if
we just move the concept of "tab_capture_end2end_tests" as a test
definition (i.e., something that only exists in the testing JSON
and as a filter file, the way site_per_process_browser_tests and
browser_side_navigation_browser_tests work today), things should
work correctly and we'll at least have one fewer way of configuring things.

TBR=kbr@chromium.org
BUG= 714336 
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel

Review-Url: https://codereview.chromium.org/2838473002
Cr-Commit-Position: refs/heads/master@{#466565}

[modify] https://crrev.com/75636899ba7c25c8b3b632989e8358987a3b0b50/chrome/test/BUILD.gn
[modify] https://crrev.com/75636899ba7c25c8b3b632989e8358987a3b0b50/content/test/gpu/generate_buildbot_json.py
[modify] https://crrev.com/75636899ba7c25c8b3b632989e8358987a3b0b50/testing/buildbot/chromium.gpu.fyi.json
[modify] https://crrev.com/75636899ba7c25c8b3b632989e8358987a3b0b50/testing/buildbot/chromium.gpu.json
[modify] https://crrev.com/75636899ba7c25c8b3b632989e8358987a3b0b50/testing/buildbot/filters/BUILD.gn
[add] https://crrev.com/75636899ba7c25c8b3b632989e8358987a3b0b50/testing/buildbot/filters/tab-capture-end2end-tests.browser_tests.filter
[modify] https://crrev.com/75636899ba7c25c8b3b632989e8358987a3b0b50/testing/buildbot/gn_isolate_map.pyl
[modify] https://crrev.com/75636899ba7c25c8b3b632989e8358987a3b0b50/tools/mb/mb.py

Comment 11 by jam@chromium.org, Apr 24 2017

Thanks Dirk for the quick fix. I verified that a reupload of https://codereview.chromium.org/2832153003/ has browser_tests running.
Project Member

Comment 12 by bugdroid1@chromium.org, Apr 24 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/eb27ad73c3ea58b8d878fa126490a3670ea98efe

commit eb27ad73c3ea58b8d878fa126490a3670ea98efe
Author: dpranke <dpranke@chromium.org>
Date: Mon Apr 24 16:04:38 2017

Revert of Rework tab_capture_end2end_tests' test definition. (patchset #2 id:40001 of https://codereview.chromium.org/2838473002/ )

Reason for revert:
Reverting, tab_capture_end2end seems to be hanging a lot of tasks. Details to
follow ...

Original issue's description:
> Rework tab_capture_end2end_tests' test definition.
>
> tab_capture_end2end_tests was previously defined as a target in
> GN that depended on browser_tests, and as a target in the gn_isolate_map
> as a "gpu_browser_test". In addition, the //testing/buildbot/ json
> file generated by //content/test/gpu/generate_buildbot_json ended up
> duplicating some of the logic that was in mb.py for handling
> gpu_browser_tests.
>
> The fact that the logic was scattered over multiple places ended up
> confusing `mb analyze`, with the result that if you were trying to
> run both browser_tests and tab_capture_end2end_tests, you might only
> get one of them.
>
> We need to reduce or eliminate the multiple different ways we have
> of configuring test steps. Eventually we should push as much of this
> logic as we can into the BUILD.gn files, I think, but for now if
> we just move the concept of "tab_capture_end2end_tests" as a test
> definition (i.e., something that only exists in the testing JSON
> and as a filter file, the way site_per_process_browser_tests and
> browser_side_navigation_browser_tests work today), things should
> work correctly and we'll at least have one fewer way of configuring things.
>
> TBR=kbr@chromium.org
> BUG= 714336 
> CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
>
> Review-Url: https://codereview.chromium.org/2838473002
> Cr-Commit-Position: refs/heads/master@{#466565}
> Committed: https://chromium.googlesource.com/chromium/src/+/75636899ba7c25c8b3b632989e8358987a3b0b50

TBR=kbr@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG= 714336 

Review-Url: https://codereview.chromium.org/2840603002
Cr-Commit-Position: refs/heads/master@{#466636}

[modify] https://crrev.com/eb27ad73c3ea58b8d878fa126490a3670ea98efe/chrome/test/BUILD.gn
[modify] https://crrev.com/eb27ad73c3ea58b8d878fa126490a3670ea98efe/content/test/gpu/generate_buildbot_json.py
[modify] https://crrev.com/eb27ad73c3ea58b8d878fa126490a3670ea98efe/testing/buildbot/chromium.gpu.fyi.json
[modify] https://crrev.com/eb27ad73c3ea58b8d878fa126490a3670ea98efe/testing/buildbot/chromium.gpu.json
[modify] https://crrev.com/eb27ad73c3ea58b8d878fa126490a3670ea98efe/testing/buildbot/filters/BUILD.gn
[delete] https://crrev.com/11bfa89469cf59b8a5792f5cbfb8f0d2ed9ae48b/testing/buildbot/filters/tab-capture-end2end-tests.browser_tests.filter
[modify] https://crrev.com/eb27ad73c3ea58b8d878fa126490a3670ea98efe/testing/buildbot/gn_isolate_map.pyl
[modify] https://crrev.com/eb27ad73c3ea58b8d878fa126490a3670ea98efe/tools/mb/mb.py

Status: Started (was: Fixed)
re-opening, apparently tab_capture is hanging the swarming bots, and so we had to revert the change and I'm landing a different CL that will just disable the test step and get browser_tests working again until we can figure out what's going on.
See bug 714542 for some of the related fallout.

Comment 15 by jam@chromium.org, Apr 24 2017

ah was just about to reopen this, now I'm not seeing browser_tests on https://codereview.chromium.org/2835233002/
Yup, that's expected until https://codereview.chromium.org/2828143007/ lands.

Comment 17 by kbr@chromium.org, Apr 24 2017

Argh. Sorry about the trouble this step is causing.

Should we temporarily comment out tab_capture_end2end_tests from the GN file as well as generate_buildbot_json.py? I can respecify the step as a pure dependency on browser_tests with the appropriate --gtest-filter arg.

Would you like me to take over this bug?

I think this is supposedly one of the logs from a bot that then became unresponsive:

https://chromium-swarm.appspot.com/task?id=35b934adcd1cf910&refresh=10&show_raw=1

Comment 19 by kbr@chromium.org, Apr 24 2017

Blockedon: 582094
Whatever was killing the Linux Swarming slaves may have been the same issue that caused content_unittests to kill the slaves in  Issue 582094 .

It looked like a signal was propagating out to the parent process.

Given that tab_capture_end2end_tests is running the same binary as browser_tests, and that the build flags shouldn't have changed when removing the notion of tab_capture_end2end_tests from the GN files, what could have changed?

maruel@ pointed at 

https://chromium-swarm.appspot.com/bot?id=build148-m4&selected=1&show_all_events=true&sort_stats=total%3Adesc

to show a list of failing tasks, but I'm not sure how to interpret them to see that.
The first (obvious) guess is that there's something in the tests themselves that is causing the problem, right?
Project Member

Comment 22 by bugdroid1@chromium.org, Apr 24 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/2b3b32f690472ea7bfb3c676adce4f3874bab95c

commit 2b3b32f690472ea7bfb3c676adce4f3874bab95c
Author: dpranke <dpranke@chromium.org>
Date: Mon Apr 24 19:00:30 2017

Attempt to disable tab_capture_end2end_tests on the trybots.

There is currently a weird interaction going on between
tab_capture_end2end_tests and browser_tests on the trybots, as the
result of some bugs in the `analyze` step. It's not immediately
obvious what the best way to fix this is, but it looks like
browser_tests isn't being run many times when it should be, and that's
bad. I think this CL will remove the collision and cause browser_tests
to run regularly, but it will cause the tab_capture tests to *not* run
on the trybots. Hopefully this is an acceptable temporary workaround,
but we need a real fix ASAP.

TBR=kbr@chromium.org, thakis@chromium.org, jam@chromium.org, enne@chromium.org
BUG= 714336 
NOTRY=true

Review-Url: https://codereview.chromium.org/2828143007
Cr-Commit-Position: refs/heads/master@{#466704}

[modify] https://crrev.com/2b3b32f690472ea7bfb3c676adce4f3874bab95c/chrome/test/BUILD.gn
[modify] https://crrev.com/2b3b32f690472ea7bfb3c676adce4f3874bab95c/testing/buildbot/gn_isolate_map.pyl
[modify] https://crrev.com/2b3b32f690472ea7bfb3c676adce4f3874bab95c/tools/mb/mb.py

Comment 23 by jam@chromium.org, Apr 24 2017

I'm seeing other problems too, i.e.
https://codereview.chromium.org/2835233002/#ps20001
should be running interactive_ui_tests and sync_integration_tests but it's not. The first patchset did run them.

Comment 24 by jam@chromium.org, Apr 24 2017

Labels: -Pri-1 Pri-0
(raising the priority since this is affecting the CQ landing untested changes)

Comment 25 by jam@chromium.org, Apr 24 2017

Can we revert the change that triggered this (https://codereview.chromium.org/2569333002) immediately?
Actually, I think the current brokenness is due to my fix -- https://codereview.chromium.org/2828143007 -- not thakis' change https://codereview.chromium.org/2569333002 .

Fixing now.
Project Member

Comment 27 by bugdroid1@chromium.org, Apr 24 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/b7b183fa716d3b18f69ebc18e290eb97d735f296

commit b7b183fa716d3b18f69ebc18e290eb97d735f296
Author: dpranke <dpranke@chromium.org>
Date: Mon Apr 24 23:50:16 2017

Fix analyze in mb.py.

As part of #466704, I changed some logic in mb.py to sort the inputs
that MB hands to GN, in an attempt to make the debugging output a
little easier to follow. It turns out that the way I did this was
broken, as there was a dependency on the ordering of the unsorted
values.

As a result, MB started remapping GN's analyze results back to the
wrong targets, and we've been running the wrong test steps since #466704
landed :(.

TBR=kbr@chromium.org, jam@chromium.org
BUG= 714336 
NOTRY=true

Review-Url: https://codereview.chromium.org/2843663002
Cr-Commit-Position: refs/heads/master@{#466828}

[modify] https://crrev.com/b7b183fa716d3b18f69ebc18e290eb97d735f296/tools/mb/mb.py

Components: -Infra>CQ Build
Labels: -Pri-0 Pri-1
I've landed a change that should fix the brokenness in #c26. I'm downgrading to P1 and will keep an eye on it. We still need to land a "proper" fix for the tab_capture problem as well.

Comment 29 by kbr@chromium.org, Apr 25 2017

Cc: hubbe@chromium.org
The probable reason the Linux Swarming bots were being killed was because https://codereview.chromium.org/2838473002 changed tab_capture_end2end_tests from running as a "gpu_browser_test" to "windowed_test_launcher".

The windowed_test_launcher test type does several operations that are invalid on the physical GPU bots, including spawning an xvfb server.

This seems to imply that it's absolutely necessary to have a separate GN target for these browser tests that are supposed to run on physical hardware.

I could refactor the browser_tests target into a common one sharable by both the main browser_tests executable and a new tab_capture_end2end_tests executable which would contain only the relevant tests. Should I embark on that or does someone have a different suggestion? The target is huge.

Update on #c28 - it looks like things are working more-or-less correctly again.
It seems like we should have a way to not have to add another GN target here. I will think about it a bit more.

Comment 32 by jam@chromium.org, Apr 25 2017

@kbr: we already have different browser test binaries, i.e. interactive_ui_tests, sync_integration_tests just in src/chrome. It should be easy to create a new one with the few tests that you need.

Comment 33 by kbr@chromium.org, Apr 25 2017

Thanks @jam. Scanned through src/chrome/test/BUILD.gn for browser_tests, interactive_ui_tests and sync_integration_tests and quickly got lost. :) It looks like these targets could use some refactoring. As one example, they all depend on sync_integration_test_support, which seems only necessary for sync_integration_tests. Will be happy to do this refactoring if I could get some guidance. I wouldn't know where to begin. Copy-pasting the existing targets seems like the wrong thing to do.

Obviously, we can add a new binary if that's good for other reasons.

However, after thinking about this a bit more, I think maybe the thing to do is to just change the wrapper scripts to support a --no-xvfb flag, and then add that flag to the JSON list of args alongside --enable-gpu.

The whole console_test_launcher/windowed_test_launcher distinction didn't really turn out to be all that useful. I think it is now only used to determine whether or not to use xvfb. For most test binaries, they can run under xvfb just fine even if they don't need X, it's just very slightly slower, and so if we get rid of the distinction I think a bunch of things will be simpler.

@kbr - you said "several operations that are invalid"; do they all stem from trying to use xvfb, or am I missing something? 

Comment 35 by kbr@chromium.org, Apr 25 2017

@dpranke I mainly noticed the xvfb bringup. Not sure what --brave-new-test-launcher does.

If xvfb.py supported --no-xvfb it would be fine by me to add it to generate_buildbot_json.py.

Comment 36 by kbr@chromium.org, Apr 25 2017

Cc: machenb...@chromium.org
Failures of the "isolate tests" step are showing up on win_optional_gpu_tests_rel. See this attempted V8 roll:
https://codereview.chromium.org/2842863002/

and for example this build:
https://build.chromium.org/p/tryserver.chromium.win/builders/win_optional_gpu_tests_rel/builds/9485

Relevant portion of the log:

----------------------
15:33:56.225452 PushDirectory(E:\b\c\b\win\src\net\tools\testserver\) = 12 files

tab_capture_end2end_tests  GetFileAttributesEx E:\b\c\b\win\src\testing\buildbot\filters\tab-capture-end2end-tests.browser_tests.filter: The system cannot find the file specified.

angle_unittests  GetFileAttributesEx E:\b\c\b\win\src\testing\buildbot\filters\tab-capture-end2end-tests.browser_tests.filter: The system cannot find the file specified.

swiftshader_unittests  GetFileAttributesEx E:\b\c\b\win\src\testing\buildbot\filters\tab-capture-end2end-tests.browser_tests.filter: The system cannot find the file specified.

angle_deqp_gles31_tests  GetFileAttributesEx E:\b\c\b\win\src\testing\buildbot\filters\tab-capture-end2end-tests.browser_tests.filter: The system cannot find the file specified.

angle_deqp_gles3_tests  GetFileAttributesEx E:\b\c\b\win\src\testing\buildbot\filters\tab-capture-end2end-tests.browser_tests.filter: The system cannot find the file specified.

angle_deqp_egl_tests  GetFileAttributesEx E:\b\c\b\win\src\testing\buildbot\filters\tab-capture-end2end-tests.browser_tests.filter: The system cannot find the file specified.

angle_deqp_gles2_tests  GetFileAttributesEx E:\b\c\b\win\src\testing\buildbot\filters\tab-capture-end2end-tests.browser_tests.filter: The system cannot find the file specified.

angle_perftests  GetFileAttributesEx E:\b\c\b\win\src\testing\buildbot\filters\tab-capture-end2end-tests.browser_tests.filter: The system cannot find the file specified.

gles2_conform_test  GetFileAttributesEx E:\b\c\b\win\src\testing\buildbot\filters\tab-capture-end2end-tests.browser_tests.filter: The system cannot find the file specified.

telemetry_gpu_integration_test  GetFileAttributesEx E:\b\c\b\win\src\testing\buildbot\filters\tab-capture-end2end-tests.browser_tests.filter: The system cannot find the file specified.
[43735] [36/798.8Mib/42025] [31/36] [0/0b/21/341.2Mib] 3s
15:33:57.425572 Uploaded 9.09Mib: out\Release\swiftshader_unittests.exe.pdb
[43735] [36/798.8Mib/42025] [31/36] [1/9.09Mib/21/341.2Mib] 4s
15:34:04.433273 Uploaded 16.3Mib: out\Release\angle_deqp_libgles31.dll.pdb
[43735] [36/798.8Mib/42025] [31/36] [2/25.4Mib/21/341.2Mib] 11s
15:34:04.899320 Uploaded 21.6Mib: out\Release\video_decode_accelerator_unittest.exe.pdb
15:34:05.003330 Uploaded 22.3Mib: out\Release\audio_unittests.exe.pdb
15:34:05.051335 Uploaded 21.0Mib: out\Release\angle_perftests.exe.pdb
15:34:05.121342 Uploaded 20.3Mib: out\Release\angle_unittests.exe.pdb
15:34:05.171347 Uploaded 20.0Mib: out\Release\angle_white_box_tests.exe.pdb
15:34:05.522382 Uploaded 26.9Mib: out\Release\gl_tests.exe.pdb
[43735] [36/798.8Mib/42025] [31/36] [8/157.5Mib/21/341.2Mib] 12s
Loading streams (has the build crashed?)...
----------------------

What do we need to do to fix the build at this point?

I thought I got rid of that file?

Comment 38 by kbr@chromium.org, Apr 26 2017

Note: same issue on Linux: https://build.chromium.org/p/tryserver.chromium.linux/builders/linux_optional_gpu_tests_rel/builds/8167 .

However a later run passed, so I wonder whether ToT is OK at this point.

Comment 39 by kbr@chromium.org, Apr 26 2017

Never mind -- the latest win_optional_gpu_tests_rel tryjob from that CL just got past the isolate_tests step.
https://build.chromium.org/p/tryserver.chromium.win/builders/win_optional_gpu_tests_rel/builds/9492

The previous tryjobs all hit the range where ToT was broken.

Labels: -Pri-1 Pri-0
This is likely causing  http://crbug.com/714976  and leads to inconsistent browser_tests and tab_capture_end2end_tests.
Blocking: 714976
Project Member

Comment 42 by bugdroid1@chromium.org, Apr 26 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/b8d34ab3bfbbc9cd34d48a72bad3fdb0ea5507b4

commit b8d34ab3bfbbc9cd34d48a72bad3fdb0ea5507b4
Author: machenbach <machenbach@chromium.org>
Date: Wed Apr 26 14:10:50 2017

Partial revert of "Attempt to disable tab_capture_end2end_tests on the trybots"

This partially reverts:
https://codereview.chromium.org/2828143007
to fix broken GPU Linux builder.

BUG= 714976 , 714336 
TBR=dpranke@chromium.org
NOTRY=true

Review-Url: https://codereview.chromium.org/2842963003
Cr-Commit-Position: refs/heads/master@{#467318}

[modify] https://crrev.com/b8d34ab3bfbbc9cd34d48a72bad3fdb0ea5507b4/testing/buildbot/gn_isolate_map.pyl

Project Member

Comment 43 by bugdroid1@chromium.org, Apr 26 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/7dad4681ebb1ef820a076c8391fa16ff1afa21c0

commit 7dad4681ebb1ef820a076c8391fa16ff1afa21c0
Author: dpranke <dpranke@chromium.org>
Date: Wed Apr 26 23:14:55 2017

Make tab_capture_end2end_tests work as a regular windowed_test_launcher.

Previously, when run under swarming, tab_capture_end2end_tests was
specified in gn_isolate_map.pyl as a 'gpu_test_launcher' rather than
as a 'windowed_test_launcher'. This, plus the way the target needed to
be declared in the BUILD.gn files was causing complications for analyze.

We think that there's no need for this separate kind of test_launcher,
as long as there is a way to tell MB to not actually launch xvfb. This
CL adds that support (by making xvfb.py be aware of a --no-xvfb flag)
and allows us to simplify a few things as a result.

Perhaps more interestingly, this'll allow us to actually get rid of the
console_test_launcher/windowed_test_launcher distinction altogether in
the future, but that's for a follow-up CL.

R=kbr@chromium.org
BUG= 714336 
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
NOTRY=true

Review-Url: https://codereview.chromium.org/2840953002
Cr-Commit-Position: refs/heads/master@{#467501}

[modify] https://crrev.com/7dad4681ebb1ef820a076c8391fa16ff1afa21c0/chrome/test/BUILD.gn
[modify] https://crrev.com/7dad4681ebb1ef820a076c8391fa16ff1afa21c0/content/test/gpu/generate_buildbot_json.py
[modify] https://crrev.com/7dad4681ebb1ef820a076c8391fa16ff1afa21c0/testing/buildbot/chromium.gpu.fyi.json
[modify] https://crrev.com/7dad4681ebb1ef820a076c8391fa16ff1afa21c0/testing/buildbot/chromium.gpu.json
[modify] https://crrev.com/7dad4681ebb1ef820a076c8391fa16ff1afa21c0/testing/buildbot/gn_isolate_map.pyl
[modify] https://crrev.com/7dad4681ebb1ef820a076c8391fa16ff1afa21c0/testing/xvfb.py
[modify] https://crrev.com/7dad4681ebb1ef820a076c8391fa16ff1afa21c0/tools/mb/mb.py

This should be fixed now. I'm going to wait for a few more builds to cycle through before closing things, though.
Status: Fixed (was: Started)
This looks more-or-less good so far, so I'm going to mark this fixed. Feel free to re-open or file a new bug if you see other issues.

Comment 46 by kbr@chromium.org, Apr 27 2017

Thanks Dirk for fixing this.

Sign in to add a comment