Closed Bug 1278402 Opened 8 years ago Closed 8 years ago

add linux64-cc platform and browser-chrome tests to taskcluster

Categories

(Testing :: General, defect)

Product:
Testing
Component:
General
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

(firefox52 fixed)

Status:
RESOLVED FIXED
Milestone:
mozilla52
Tracking Flags:
Tracking Status
firefox52 --- fixed

People

(Reporter: jmaher, Unassigned)

References

(Blocks 1 open bug)

Details

Attachments

(4 files)

Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.
58 bytes, text/x-review-board-request
jmaher
: review+
dustin
: review+
Details
After patch for | try: -b o -p all -u none -t none |
1.12 KB, application/text
Details
Before patch for | try: -b o -p all -u none -t none |
1.12 KB, application/text
Details
Screenshot from 2016-09-21 09-57-26.png
34.78 KB, image/png
Details 
currently there is an unused platform (linux64-cc) in buildbot to compile code coverage (-ccov).  In fact this doesn't work today as compiler tools have upgraded.

For now, lets focus on adding/editing .yml files to get linux64-cc as a platform and run the mochitest-browser-chrome tests in opt mode as a starting point.
Blocks: 1278404
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review72610

::: taskcluster/ci/desktop-test/test-platforms.yml:37
(Diff revision 1)
> +    build-platform: linux64-ccov/debug
> +    test-set: ccov-code-coverage-tests
> +linux64-jsdcov/opt:
> +    build-platform: linux64-jsdcov/opt
> +    test-set: jsdcov-code-coverage-tests
> +linux64-jsdcov/debug:

I really think we should stick with opt for now- I know we run a few different tests and build with some extra assertions in debug- it seems expensive for the small delta in coverage.

::: taskcluster/ci/desktop-test/tests.yml:218
(Diff revision 1)
>      allow-software-gl-layers: false
>  
> +mochitest-browser-chrome-jsdcov:
> +    description: "Mochitest browser-chrome jsdebugger code coverage run"
> +    suite: mochitest/browser-chrome-coverage
> +    treeherder-symbol: tc-MCov(bc)

leave this as tc-M(bc)

::: taskcluster/ci/desktop-test/tests.yml:224
(Diff revision 1)
> +    loopback-video: true
> +    chunks: 35
> +    e10s: false
> +    max-run-time:
> +        by-test-platform:
> +            default: 12000

do we need the 12000 (3+ hours) for 35 chunks?

::: taskcluster/ci/desktop-test/tests.yml:242
(Diff revision 1)
> +    allow-software-gl-layers: false
> +
> +mochitest-browser-chrome-ccov:
> +    description: "Mochitest browser-chrome jsvm code coverage run"
> +    suite: mochitest/browser-chrome-chunked
> +    treeherder-symbol: tc-MCov(bc)

lets leave this as tc-M(bc)

::: taskcluster/ci/desktop-test/tests.yml:248
(Diff revision 1)
> +    loopback-video: true
> +    chunks: 7
> +    e10s: false
> +    max-run-time:
> +        by-test-platform:
> +            default: 12000

do we need 12000 (3+ hours) for this run?  I understand it runs slower, but this seems like a lot slower.

one thing to consider here is using the existing mochitest-browser-chrome and doing:
max-run-time:
    by-test-platform:
        default: xxxx
        linux64-ccov: 7200 (or whatever is really needed)
mozharness:
    extra-options:
        by-test-platform:
            linux64-ccov:
                - --code-coverage
                - --mochitest-suite=browser-chrome-chunked
            default:
                - --mochitest-suite=browser-chrome-chunked


this would reduce the test sets- either we can do this for both types (jsdcov and ccov), or make a coverage item for tests and do a by-test-platform for the differences.  I would like to have as little as possible test sets added.

::: taskcluster/taskgraph/transforms/make_task.py:177
(Diff revision 1)
>      'tc-Fxfn-l': 'Firefox functional tests (local) executed by TaskCluster',
>      'tc-Fxfn-l-e10s': 'Firefox functional tests (local) executed by TaskCluster with e10s',
>      'tc-Fxfn-r': 'Firefox functional tests (remote) executed by TaskCluster',
>      'tc-Fxfn-r-e10s': 'Firefox functional tests (remote) executed by TaskCluster with e10s',
>      'tc-M': 'Mochitests executed by TaskCluster',
> +    'tc-MCov': 'Mochitests code coverage executed by TaskCluster',

please remove this line

::: testing/mochitest/runtests.py:2437
(Diff revision 1)
>              self.stopServers()
>  
> +        ignoreMissingLeaks = options.ignoreMissingLeaks
> +        leakThresholds = options.leakThresholds
> +
> +        if options.jscov_dir_prefix:

while this is "self explanatory", I would like to have a comment indicating why this is necessary.
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review72612

overall, this is a great starting point- I asked for a few specific things and asked a few questions and general things we should consider doing.
Attachment #8785602 - Flags: review?(jmaher) → review-
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review72610

> do we need 12000 (3+ hours) for this run?  I understand it runs slower, but this seems like a lot slower.
> 
> one thing to consider here is using the existing mochitest-browser-chrome and doing:
> max-run-time:
>     by-test-platform:
>         default: xxxx
>         linux64-ccov: 7200 (or whatever is really needed)
> mozharness:
>     extra-options:
>         by-test-platform:
>             linux64-ccov:
>                 - --code-coverage
>                 - --mochitest-suite=browser-chrome-chunked
>             default:
>                 - --mochitest-suite=browser-chrome-chunked
> 
> 
> this would reduce the test sets- either we can do this for both types (jsdcov and ccov), or make a coverage item for tests and do a by-test-platform for the differences.  I would like to have as little as possible test sets added.

Unfortunatly, we can't do this. The 'extra-options' field cannot be defined y using the 'by-test-platform' option. Take a look at this file, it shows the schema: https://dxr.mozilla.org/mozilla-central/source/taskcluster/taskgraph/transforms/tests/test_description.py?q=test_description.py&redirect_type=direct .

I will reduce the time to 7200 though as I think that is all we need at the most based on the tests I just ran (with 'bc35' usually being the longest).
possibly jsdcov could be by-test-platform then :)
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review72610

> Unfortunatly, we can't do this. The 'extra-options' field cannot be defined y using the 'by-test-platform' option. Take a look at this file, it shows the schema: https://dxr.mozilla.org/mozilla-central/source/taskcluster/taskgraph/transforms/tests/test_description.py?q=test_description.py&redirect_type=direct .
> 
> I will reduce the time to 7200 though as I think that is all we need at the most based on the tests I just ran (with 'bc35' usually being the longest).

I can't do jsdcov either. The 'suite' field is missing the 'by-test-platform' option.
(In reply to Greg Mierzwinski from comment #8)
> https://treeherder.mozilla.org/#/jobs?repo=try&revision=40ff15a12e09

This push shows the 'tc-MCov' treeherder symbol but it is no longer in the code so in subsequent pushes it should be (hopefully) gone.
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review72750

thank you for the cleanup of comments and removing the debug bits.  Overall this is looking better.  I still don't like the extra test types, but I see that we would need by-test-platform support for both suite and extra-options.  I would like to get a bug on file (good for the upcoming UCOSP program to add support for that or find another creative way to reduce this (possibly with a custom transform: https://dxr.mozilla.org/mozilla-central/source/taskcluster/taskgraph/task/transform.py).  Getting this landed lets us move onto new and exciting things!
Attachment #8785602 - Flags: review?(jmaher) → review+
See Also: → 890116
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

:dustin can you look at this patch and ensure that it is only going to run on try, not on other branches?  I am having a brain fart on in-tree configs today.
Attachment #8785602 - Flags: feedback?(dustin)
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review75646

I'm really excited to see the self-serve design supporting the introduction of new tasks like this!

I have a few comments on the patch itself, and then some bad news :(

(In reply to Greg Mierzwinski from comment #5)
> Unfortunatly, we can't do this. The 'extra-options' field cannot be defined
> y using the 'by-test-platform' option. Take a look at this file, it shows
> the schema:

The schema's not written in stone -- you could certainly add such an option if it's useful to you!  In general, you should not feel like the implementation is a barrier here, any more than you might feel like an existing class's method signatures can never be changed.  If a change is well-designed and makes sense, make it!

::: taskcluster/ci/desktop-test/tests.yml:256
(Diff revision 2)
> +        no-read-buildbot-config: true
> +        config:
> +            by-test-platform:
> +                default:
> +                    - mozharness/configs/unittests/linux_unittest.py
> +                    - mozharness/configs/remove_executables.py

If there's only a "default" option, then there's no need for by-test-platform (applies here and several other places)

::: taskcluster/ci/legacy/tasks/branches/base_job_flags.yml:26
(Diff revision 2)
>      - linux
>      - linux-l10n    # Desktop l10n
>      - linux64
>      - linux64-l10n  # Desktop l10n
> +    - linux64-ccov
> +    - linux64-jsdcov

Adding these here will make the tasks run on all branches.

I agree with Joel that introducing new test names is not great.  I can see two paths to fixing this, depending on where you see this development going in the future.

First, we could introduce a new kind specifically for this purpose.  If you expect the coverage measurement to grow in a different direction than the suite of "regular" tests, that might make sense.

The other option is to "split" tasks by coverage, similar to what we do for e10s.  You'll have noticed that if the `e10s` property is "both", then two tasks are created -- one with, and one without, e10s enabled.  You could introduce a similar split for coverage:

 * coverage: no --> only run the test without coverage (default)
 * coverage: yes --> run the test without coverage and with each coverage type
 * coverage: jdscov --> only run with jsdcov
 * coverage: [no, ccov] --> run without coverage and with ccov

This would make a lot of sense if the intent is to eventually have a portion of the test suite running both with and without coverage (noting, of course, that that could be a SUBSTANTIAL additioal testing load -- we already run compute-days of tests for each push, and this could potentially double or triple that load).  This will also be a little tricky since it appears from your patch that these coverage options require distinct installers.

OK, now the bad news.  In bug 1286075, I'm refactoring how all of the builds are done, with the side-effect of deleting everything under `taskcluser/ci/legacy`.  This has been a months-long process and is just about to go up for its second round of review.  So while you're pondering and refactoring based on the comments above, please also plan to  rebase on top of that patchset.  That will affect the implementation of the coverage builds, but not the tests.
Attachment #8785602 - Flags: review-
Attachment #8785602 - Flags: feedback?(dustin)
Bug 1286075 should be landing in a day or two -- final review is up and I don't expect any last-minute changes.
(In reply to Dustin J. Mitchell [:dustin] from comment #14)
> Bug 1286075 should be landing in a day or two -- final review is up and I
> don't expect any last-minute changes.

Thanks Dustin!
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review75646

> If there's only a "default" option, then there's no need for by-test-platform (applies here and several other places)

Should I clean-up the other tests as well since most of them have this in their mozharness config option?
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review75646

> Should I clean-up the other tests as well since most of them have this in their mozharness config option?

I'm sorry, no, they can be left including in your test stanza.  Geoff Brown added them in anticipation of also supporting windows and MacOS X tests with these same definitions.
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review77518

the mochitest/mozharness bits still look great, and I am happy to see by-test-platform being used to simplify the taskcluster configs- my only question would be related to does this run on try server only?

::: taskcluster/ci/desktop-test/tests.yml:196
(Diff revision 3)
>  mochitest-browser-chrome:
> -    description: "Mochitest browser-chrome run"
> -    suite: mochitest/browser-chrome-chunked
> +    description:
> +        by-test-platform:
> +            linux64-jsdcov/opt: "Mochitest browser-chrome JSDebugger code coverage run"
> +            linux64-ccov/opt: "Mochitest browser-chrome GCOV code coverage run"
> +            default: "Mochitest browser-chrome run"

this might be overkill, but it is cool to see that we can make this more descriptive :)
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review77578

Nice work!

::: taskcluster/ci/desktop-test/tests.yml:205
(Diff revision 3)
> +            default: mochitest/browser-chrome-chunked
>      treeherder-symbol: tc-M(bc)
>      loopback-video: true
> -    chunks: 7
> +    chunks:
> +        by-test-platform:
> +            linux64-jsdcov/opt: 35

wow, that's a lot of chunks! :)
Attachment #8785602 - Flags: review?(dustin) → review+
Pushed by jmaher@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/9c949b216b37
Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster. r=dustin,jmaher
thanks for backing this out :kwierso.

This was running jobs on all pushes- we want this on Try server only, eventually on the nightly scheduler.

Here is a view of what we see on autoland:
https://treeherder.mozilla.org/#/jobs?repo=autoland&revision=076af765a26fdf04c055ae525af11559a4a745d7&filter-resultStatus=success&filter-resultStatus=testfailed&filter-resultStatus=busted&filter-resultStatus=exception&filter-resultStatus=retry&filter-resultStatus=running&filter-resultStatus=pending&filter-resultStatus=runnable&filter-searchStr=cov

2 things:
1) this shouldn't be running by default on autoland!
2) the build and test lines end up with slightly different descriptions and clutter up treeherder, lets make sure they are referencing the same platform.
* build is: Linux x64 ccov
* tests are: linux64-ccov opt

we should be able to fix that both for ccov and jsdcov.

I am not sure how to make this run on try only, :greg, can you take a look at this- possibly :dustin can help us if we get stuck with making this try only (and hopefully not scheduled with -p all as well)
Flags: needinfo?(jmaher) → needinfo?(gmierz2)
(In reply to Joel Maher ( :jmaher) from comment #24)
> thanks for backing this out :kwierso.
> 
> This was running jobs on all pushes- we want this on Try server only,
> eventually on the nightly scheduler.
> 
> Here is a view of what we see on autoland:
> https://treeherder.mozilla.org/#/
> jobs?repo=autoland&revision=076af765a26fdf04c055ae525af11559a4a745d7&filter-
> resultStatus=success&filter-resultStatus=testfailed&filter-
> resultStatus=busted&filter-resultStatus=exception&filter-
> resultStatus=retry&filter-resultStatus=running&filter-
> resultStatus=pending&filter-resultStatus=runnable&filter-searchStr=cov
> 
> 2 things:
> 1) this shouldn't be running by default on autoland!
> 2) the build and test lines end up with slightly different descriptions and
> clutter up treeherder, lets make sure they are referencing the same platform.
> * build is: Linux x64 ccov
> * tests are: linux64-ccov opt
> 
> we should be able to fix that both for ccov and jsdcov.
> 
> I am not sure how to make this run on try only, :greg, can you take a look
> at this- possibly :dustin can help us if we get stuck with making this try
> only (and hopefully not scheduled with -p all as well)

I've been looking around and found that we can use the 'run-on-projects' flag set as an empty list in linux.yml to make it run only on try and not run with '-p all'. Unfortunately, it seems like the flag isn't being recognized or something. Here are the changes that I've made (they are in linux.yml): https://hg.mozilla.org/try/diff/014691f84364/taskcluster/ci/build/linux.yml .

Here's a test run with '-p all': https://treeherder.mozilla.org/#/jobs?repo=try&revision=8d9bfa563de882286fc0ceae86e494cfb6eb4b1d .

It is still being scheduled when it shouldn't be based on my understanding of how it works, but I think I'm missing something somewhere. Also if you look at 'task_graph.json', you'll see that "run_on_projects" is always 'all' for jsdcov or ccov.

Dustin, would you know if I am missing something here? I am thinking of adding it into the RIDEALONG_BUILDS list to see if that solves it.
Flags: needinfo?(gmierz2) → needinfo?(dustin)
RIDEALONG_BUILDS is a way to get things run on try that otherwise wouldn't, which is not what you want here -- you want things run less frequently on try than they otherwise would be.

Note that you can simulate these runs locally, rather than pushing to try and cancelling, using the `./mach taskgraph` subcommands.

If you look in full-task-graph.json, you'll see that the build jobs have run-on-project: [] as expected, but the test jobs that depend on it have run-on-project: ['all'].  That means that the test jobs are scheduled to run on try, and since they depend on the build job it gets pulled along for the ride.

So to fix, you'll need to have run-on-project: [] set for all of those test jobs, too.
Flags: needinfo?(dustin)
Attachment #8785602 - Flags: review+ → review?
Attachment #8785602 - Flags: review?(dustin)
Thanks for the help :dustin! I'm pretty sure I've solved the problem but I'm not sure if it's the correct way of doing it or if there is another way of doing it. Here's a test run with this patch that was just submitted for review:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=16626091fe15c49d07ae00fc28b1824d803e8d8f
Greg, can you verify that -p all yields no jsdcov or ccov?

what :dustin recommended by running |mach taskgraph| is the best way to do this.  you can download parameters.yml from your previous try push as an artifact of the gecko decision task:
https://queue.taskcluster.net/v1/task/EstObAG3QkOo9egqoJnKwQ/runs/0/artifacts/public%2Fparameters.yml

you could edit that to make it |-u none| which would do just builds.

then finally run:
./mach taskgraph target -p <path>/parameters.yml

maybe before/after your patch to find parity?
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review78490

Sorry to switch back to `r-` -- but at least it's just the one issue!

::: taskcluster/ci/desktop-test/tests.yml:196
(Diff revision 4)
>  mochitest-browser-chrome:
> -    description: "Mochitest browser-chrome run"
> -    suite: mochitest/browser-chrome-chunked
> +    description:
> +        by-test-platform:
> +            linux64-jsdcov/opt: "Mochitest browser-chrome JSDebugger code coverage run"
> +            linux64-ccov/opt: "Mochitest browser-chrome GCOV code coverage run"
> +            default: "Mochitest browser-chrome run"

Joel mentioned this might be overkill.  It does seem a bit verbose (why not have a case for all of the platforms?).  Is there a strong reason to include these distinct descriptions for these two jobs?

::: taskcluster/taskgraph/transforms/task.py:524
(Diff revision 4)
> +        if attributes['build_platform'] in ['linux64-ccov', 'linux64-jsdcov']:
> +            attributes['run_on_projects'] = []
> +

This is too purpose-specific for this generic transform -- we shouldn't have anything platform-specific here.  Instead, this should be specified in the test description (`tests.yml`) and threaded through to so that it arrives intact at `task['run-on-projects']`.  That would mean adding the attribute to `taskcluster/taskgraph/transforms/tests/test_description.py` and copying it along, if present in `taskcluster/taskgraph/transforms/tests/make_task_description.py`.
Attachment #8785602 - Flags: review?(dustin) → review-
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review78556

::: taskcluster/ci/build/linux.yml:181
(Diff revision 4)
>  
> +linux64-jsdcov/opt:
> +    description: "Linux64-JSDCov Opt"
> +    index:
> +        product: firefox
> +        job-name: Linux64-JSDCov-opt

Sorry I missed commenting on this earlier.  Our job names are lowercase.  I think the name looks good otherwise (it's generally `{platform}-{type}`).
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review78490

> Joel mentioned this might be overkill.  It does seem a bit verbose (why not have a case for all of the platforms?).  Is there a strong reason to include these distinct descriptions for these two jobs?

I just thought that a description of the difference would be suitable but it is definitely verbose. I'll remove this and the schema changes that I made for this.

> This is too purpose-specific for this generic transform -- we shouldn't have anything platform-specific here.  Instead, this should be specified in the test description (`tests.yml`) and threaded through to so that it arrives intact at `task['run-on-projects']`.  That would mean adding the attribute to `taskcluster/taskgraph/transforms/tests/test_description.py` and copying it along, if present in `taskcluster/taskgraph/transforms/tests/make_task_description.py`.

Thank you for the guidance! I wasn't sure of the best way to do this so I'll add an optional run-on-projects flag to the tests schema and run it through.
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review78556

> Sorry I missed commenting on this earlier.  Our job names are lowercase.  I think the name looks good otherwise (it's generally `{platform}-{type}`).

I'll change that. I was trying to make the build name similar to the test names that are shown on treeherder: https://treeherder.mozilla.org/#/jobs?repo=try&revision=16626091fe15c49d07ae00fc28b1824d803e8d8f .
These are the target tasks produced after the patch was applied.
These are the target tasks produced before the patch.
Attachment #8793125 - Attachment mime type: application/json → application/text
Attachment #8793127 - Attachment mime type: application/json → application/text
ok, these target tasks are the same!
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review78732

I think these small changes will allow this to display the tests on the same line as the build in treeherder.

::: taskcluster/ci/build/linux.yml:178
(Diff revision 5)
>          custom-build-variant-cfg: asan-tc-and-debug
>          tooltool-downloads: public
>          need-xvfb: true
>  
> +linux64-jsdcov/opt:
> +    description: "Linux64-JSDCov Opt"

posibly we need to change this to be "Linux x64 JSDCov Opt"

::: taskcluster/ci/build/linux.yml:203
(Diff revision 5)
> +        secrets: true
> +        tooltool-downloads: public
> +        need-xvfb: true
> +
> +linux64-ccov/opt:
> +    description: "Linux64-CCov Opt"

posibly we need to change this to be "Linux x64 CCov Opt"
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review78732

As far as I understand, the description is not used to determine anything in Treeherder.  It is displayed as the "Summary" under "Job Details", but that's about it. It's not even used to create the "Job (sig)".  So I don't think it really matters what the description is, aside from aesthetics.
Here's an example of where the description shows up
so I might have been wrong on my assertion to change the name
Comment on attachment 8785602 [details]
Bug 1278402 - Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster.

https://reviewboard.mozilla.org/r/74754/#review78846
Attachment #8785602 - Flags: review?(dustin) → review+
Pushed by jmaher@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/2fbbf102636a
Add linux64-ccov and linux64-jsdcov as code coverage build platforms to taskcluster. r=dustin,jmaher
https://hg.mozilla.org/mozilla-central/rev/2fbbf102636a
Status: NEW → RESOLVED
Closed: 8 years ago
status-firefox52: --- → fixed
Resolution: --- → FIXED
Target Milestone: --- → mozilla52
Do these get run regularly anywhere yet? Where does the coverage data go afterwards?
historically these have never been run regularly, so as it stands they are not run regularly today.

The results go to active data, we have a current project underway where we are expanding the coverage and making a system to display the results including deltas.
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: